RF Adaptive Pre-Distortion In The PALM Of Your Hand
By Paul Kruczkowski, Editor
Highly linear performance is an essential requirement for power amplifiers (PAs) applied to amplitude or multicarrier modulation schemes that are often used in video applications. Poor linearity distorts the video signal and results in unwanted power spilling into the adjacent channel, which is referred to as spectral regrowth or “shoulders.” A straightforward method to achieve high linearity is to oversize the PA, so that the amplifier is operated well below its saturated power level (PSAT), or “backed-off.” In addition to the increased size, cost, and power consumption associated with the larger amplifier, running backed-off is very inefficient. These drawbacks have led to the development of predistortion techniques that improve PA linearity with reduced back-offs, providing improved efficiency and in some cases allowing the use of lower-power parts to achieve performance goals.
Predistortion — distorting the input signal of a PA to counter how the PA alters the signal to minimize the difference between the amplified output signal and the original undistorted input signal — can be accomplished at radio frequencies (RF) or digitally. Although RF predistortion (RFPD) preceded digital predistortion (DPD), the advent of higher-speed digital signal processing ADCs and DACs made DPD the preferred method for most video applications, because the power consumption of the DPD engine and the large footprint of the complex DPD circuit was inconsequential . One limitation of DPD is that the downconversion and upconversion processes require filtering and adds delay that ultimately limits the DPD correction bandwidth.
On the other hand, for video applications where size and power consumption are significant factors, such as unmanned aerial systems (UAS), RFPD is the better choice. The linearization functions needed to predistort the signal are essentially the same for DPD and RFPD. However, the independent gain and phase adjustments in RFPD are analog and are determined by comparing the envelopes of the input signal with the PA’s output signal. RFPD circuits are more compact, have low power consumption, and have improved correction bandwidth over its DPD counterpart. Care does have to be taken control DC bias changes and to mitigate effects of temperature change.
I was recently introduced to a new RFPD solution, the Power Amplifier Linearizer Module (PALM) from NuWaves Engineering. The fully adaptive PALM is very compact, measuring only 2.50” x 1.75” x 0.75” and weighing 2.5 ounces, with a power consumption under 1 W (0.2A @ +5VDC). The first release will operate from 2200 to 2600 MHz and will provide between 12 and 20 dB improvement on ACPR, depending on the modulation and signal bandwidth.
I was impressed with the performance PALM demonstrated on a PA with Psat of 12 W amplifying a 32 QAM signal with 6.9 dB PAR that had an EVM of 3.8% and MER of 23.3 dB when operating at Psat. When the PA was backed-off 3 dB to 6 W output power and the PALM was implemented, the shoulder level improved by 12.7 dB, the EVM was 0.95%, and the MER was 35.2 dB. Assuming a 1:1 ratio for shoulder level improvement to back-off, to achieve the same performance at 6 W without PALM predistortion would require a much larger PA (near 100 W) backed way off.
The PALM will likely be used in applications that include UAS, RF communications, software defined radios (SDR), and broadband telemetry. NuWaves also plans to offer the PALM in other frequency bands between 168 MHz and 4.2 GHz, as applications arise.
I’m no stranger to fully adaptive predistortion techniques since I was an engineer in the broadcast TV industry for many years. Adaptive DPD was commonly used with class AB LDMOS PAs to increase linearity in order to improve transmitter performance and meet FCC mask requirements. It was nice to finally be exposed to DPD’s power sipping, compact predecessor, RFPD.