Targeting radar with 150 V RF GaN HEMTs
BY GABRIELE FORMICONE, FOUAD BOUERI, JEFF BURGER, JAMES CUSTER AND JOHN WALKER FROM INTEGRA TECHNOLOGIES
There is plenty of room for improvement in the high-power radar systems operating in the UHF band. Lying at the heart of the majority of these systems are travelling-wave tubes, which are bulky, fragile and have an efficiency that is limited to about 65 percent. Replacing these tubes with solid-state devices could increase efficiency and robustness, and also enable a number of other improvements, including: the introduction of lower voltage power supplies, a move to a more flexible, modular design that is easier to maintain, and a reduction in long-term system costs.
Although there are several solid-state technologies capable of delivering an output in the microwave region, most are not that attractive for deployment in high-power radar. Silicon devices, such as the bipolar junction transistor and the vertical and lateral diffused MOSFET, are used in several radars from VHF to S-band and have an output power capability that is somewhat limited. Meanwhile, attempts to take significant market share with the SiC static induction transistor have failed, due to a gain of less than 10 dB and an efficiency of barely 50 percent at 450 MHz.
A far more promising technology is the GaN-based HEMT. In general, its transmit power is held back by its operating voltage, which is typically 50 V or below up to S-band. However, thanks to sophisticated drain engineering technology developed by our team at Integra Technologies of El Segundo, California, operating voltages can reach 150 V. This leads to a dramatic improvement in output characteristics. Pulsed power densities can reach 30 W/mm, equipping radar manufacturers with a solid-state device for producing an output power of several tens of kilowatts. What's more, our HEMTs provide unprecedented gain and efficiency, combined with good reliability.
These devices are a part of our portfolio that includes more common products operating at power supply voltages of 28 V and 50 V. By supplementing these more standard offerings with variants operating at higher voltages, we are setting ourselves apart from our peers and expanding the markets that we serve "“ although we are highlighting radar, high-power RF amplifiers operating at high voltages can also be used for applications in particle accelerators, microwave sintering, as vacuum tubes replacement in general, and other applications involving industrial, scientific and medical radio bands.
Device development and production takes place at our 6-inch all-gold metallization silicon wafer fab, which includes equipment re-tooled to process 4-inch GaN-on-SiC wafers. By adopting this approach, as our GaN sales grow, we can meet the need for increased production by switching production to 6-inch wafers. This transition will double capacity, while trimming production costs.
At our headquarters, we have a fully owned 22,000 square-foot facility with a clean room. To further support design, we have another 4,000 square-foot of space in research centres in other locations.
To eliminate the risk of failing to fulfil customer orders, we employ a dual source strategy, working with domestic foundries and partners. We avoid outsourcing all production, however, because we believe that by owning a fab, we are better positioned to support customers for the expected operational lifetime of the devices, which is in excess of 30 years. During this time, supplies can be disrupted through mergers, acquisitions, natural disasters and unforeseen eventualities.
Figure 1: Saturated output power and terminal load impedance of a 15 mm, high-voltage GaN transistor. On the right hand side is the corresponding load impedance the transistor requires for optimum match. At 150 V, the 450 W device has a load impedance of 25 ohms, which is straightforward to match to 50 ohms. This medium-size transistor has an array of 60 gate fingers, each with a length of 250 mm. This die, which has dimensions of 3.9 mm by 1.0 mm, has been optimized to provide stable and reliable operation.
Increasing the powerAt our facility we have demonstrated the virtues of our high-voltage HEMTs by measuring the increase in output power with bias voltage. Using an operating frequency of 430 MHz, a pulse width of 100 ms and a 10 percent duty cycle, we have found that operation at 75 V, 100 V, 125 V and 150 V produces a saturated output power of 150 W, 250 W, 350 W and 450 W, respectively (see Figure 1).
What is particularly pleasing is that the benefits of an additional 25 V of bias are not limited to a 100 W hike in the saturated output power. Instead, they extend to an increase in the load impedance, so that it is closer to that needed to deliver maximum output power to a 50 ohm load, such as an antenna. Note that the load impedance of all RF power transistors, regardless of the materials they are made from, is relatively low, with efforts to increase the output power via an increase in gate periphery paying the expense of a lower load impedance. The smaller it is, the larger the impedance transformation required to match it to 50 ohms "“ and the greater the output loss and reduction in efficiency.
15 mm 150 V GaN transistor, with no internal output match and a LCL input match. The internal output match is not used so that optimum fundamental and harmonic impedances can be adjusted outside the device package.
The substantial success that we have had begs the question: given that a higher operating voltage produces the dual benefit of a higher output power and a higher load impedance, why is this approach not standard within industry?
The answer is that to operate at a higher voltage, the transistor must withstand a higher breakdown voltage. And to realise this and avoid catastrophic failure, the drift region of the device must be extended. Since this is the most critical region of a high-voltage RF transistor, this modification, which is known within the industry as drain engineering, is challenging and must be carried out with great care.
Even optimal drain engineering produces an increase in the transistor's on-resistance, which translates to a higher loss and diminished radar efficiency. However, this loss in efficiency is minimised with our state-of-the-art drain engineering and field plate techniques. Combining these two results in a high breakdown voltage, while minimizing impact on the on-resistance.
With AlGaN/GaN HEMTs there are more options than there are with silicon technologies, such as LDMOS, for increasing the breakdown voltage while making minimal impact on the on-resistance. One avenue, which we have taken, is to increase the breakdown voltage by adding iron into the buffer layer "“ a similar response results from a small incorporation of aluminium.
When adopting either of these approaches the key is to not be too aggressive and impair reliability. If aluminium or iron is added improperly, although on-resistance will appear to be low under DC operating tests, it will increase dynamically under RF operation; this process in GaN devices is called DC-RF dispersion, and it is very detrimental to RF device performance.
To avoid these issues, comprehensive optimisation must take place that considers the drift region length, locations of field plates, and the iron and aluminium content in the device buffer layer. Get this just right and it is possible to minimize the transistor on-resistance, realise acceptable RF performance and enable operation at a targeted voltage.
As well as producing very high output powers, our devices excel in efficiency. This exceeds 70 percent at voltages ranging from 75 V to 150 V, and peaks at 78 percent at 100 V (see Figure 2 for details). The reason why the efficiency is highest at this voltage is that the test fixture had been used for a previous project, where harmonic tuning techniques were used to optimize power at 100 V bias. Retuning the test fixture harmonic impedances will enable the efficiency at 150 V to get close to 80 percent. This adjustment is guaranteed to pay dividends, as it has already succeeded with devices operating at 75 V.
Turning to gate pulsing, a mode of operation often used for radar, can deliver a further efficiency gain of 5 percentage points on average. This can propel the efficiency of our 450 W devices, which have a gate periphery of just 15 mm, to over 80 percent at 150 V. Operating in this manner, gain is well above 22 dB, indicating that our 150 V GaN technology promises to perform well at higher UHF frequencies, as well as in the L-band, where it could serve pulse radar.
In comparison, a SiC static induction transistor operating at the same frequency produces just 10 dB of gain, and struggles to realise 50 percent drain efficiency. The lower gain has unwanted consequences: an additional transistor is needed to drive the output stage, leading to a circuit with a larger footprint and an even lower efficiency at the system level.
The route to realising even higher powers, which might be needed for applications such as pulsed radar, is to increase the size of the gate periphery.
Given the high initial impedance, lowering this is not expected to have any major drawbacks. (Note that we already have experience of devices with a larger gate periphery, as we have used more traditional 50 V GaN transistors with a 36 mm gate periphery in a multi-chip package that produces a S-band output power of 1 kW. In the L-band, we also use larger chips, which have a 50 mm gate periphery.)
If we increase the gate periphery of our 150 V device from 15 mm to around 40 mm, this will create a single chip delivering a saturated output power in excess of 1 kW. Although load impedance will fall from scaling, it would still be just over 11 ohms, which is very manageable from an impedance matching perspective. There is also the opportunity to form a four-chip, 4 kW device in a single-ended ceramic package with an output load impedance of around 3 ohms, and a six-chip, 6 kW variant in a dual lead package with a load impedance of around 2 ohms.
As well as considering the impedance of these devices, which at these power levels are in a very comfortable region from an impedance matching perspective, we must consider issues related to thermal management. For our 15 mm device driven with 100 ms pulses and a 10 percent duty cycle, according to our model, thermal resistance is 0.5°C/W at a case temperature of 26 °C (it is 2°C/W in CW operation, which is not recommended). When operating at 80 percent efficiency and producing an output of 1 kW, the DC power supplied to the die is 1.25 kW, so it dissipates 250 W. This would lead to a peak junction temperature during the 100 ms pulse of 150°C, which is well within the regime of reliable operation. In fact, the temperature will be below this, because the length of the fingers must be extended for a 1kW die design, and this modification drives down thermal resistance.
Figure 2: Drain efficiency at saturated output power for a 15 mm, high-voltage GaN transistor. The efficiency is calculated based on average current during the 100 µs pulse with a 1 ms period (or 10 percent duty cycle); in radar operation it is common to shut-off the device when the pulse is off through a gating circuit. This procedure has the effect of increasing the efficiency values reported above by an average of 5 percentage points.
To ensure high reliability, our chips are housed in a nickel- and gold-plated package that is hermetically sealed with a gold-tin solder. Gold bond wires, formed with a highly precise, automated machine, enable the highest levels of repeatability and consistency in RF performance. We have no doubt that our 150 V transistors set a new benchmark for the performance of solid-state devices for ultra-high power pulse radars in the UHF band. Preliminary data suggests that a single device with an output power in excess of 5 kW can provide 20 dB of gain and a drain efficiency of more than 70 percent.
The load impedance for this device, which would contain six larger chips and be housed in a dual lead industry-standard package, is higher than 2 ohms, making it easy to match to 50 ohms. Customers can obtain customised variants of this technology to meet their particular requirements, and a greater range of options should be available in future, as we expand our products to a wider range of frequency bands.
