Very low noise preamp (0.4 dB NF) for 1296 MHz & 1420 MHz

Created: Dec. 2021. Minor update: Mar. 2023.

Introduction

The compact (2 x 2 x 0.75mm) 8-pin quad flat non-lead (QFN) packaged microwave monolithic integrated circuit (MMIC) consists of a common-source amplifier and an active bias regulator. Its 0.25 µm feature-size GaAs enhancement-mode, pseudo-morphic high electron mobility transistor (ePHEMT) process has a high gain-product bandwidth, fT >30 GHz, that allows the target gain to be achieved in one stage. This design can simultaneously cover the amateur 1296 MHz and radio astronomy 1420 MHz bands.

Material & method

A minimal number of external components (C1-L1 and C2-L2) provided the matching and biasing functions that were not feasible to integrate at the chip level (Fig. 1). In addition to DC blocking and RF choking functions, C1-L1 also roll off undesirable gain below the operating frequency (f0). Both L1 and L2 should be operated below their self-resonant frequency (SRF) for effective choking. C3-C5 decouple RF from the bias lines. The external components are 0402-size. 

The evaluation board nominal DC bias values are Vdd = 5V, and Idd = 53 mA (set by RBIAS = 6.8kΩ). The combination of ePHEMT and high operating current enables immunity against interference. 

The 21.5 x 18 mm PCB is made of 10 mil Rogers RO4350 (fig. 3). The 2-pin connector on the top edge of the PCB is for connection to a 5V supply. 

Results

At 1296 MHz, the gain is ~15.0 dB and Input Return Loss IRL / output return loss ORL are better than -15 dB (fig. 5, blue traces).  At 1420 MHz, gain is ~14.5 dB and IRL / ORL are better than -15 dB. The simulated gain (blue dashes) agrees almost perfectly with measurement (blue solid). The simulated IRL (black dashes) has less than 4 dB error over 1.1 ~ 1.5 GHz. The predicted ORL has a maximum error of 9 dB, but this is not a cause for concern because the matching is good enough. The most probable cause of these discrepancies is error in modelling the passive components and PCB. 

Measured noise figure (F) is lower than 0.5 dB at 1296 and 1400 MHz (fig. 6). Although the evaluation ends at 1400 MHz, the 1420 MHz's F is not expected to differ significantly. The simulated F is ~ 0.1 dB higher than experimental - the most probable cause of this (insignificant) discrepancy is error in modelling the passive components and PCB.  It may be possible to reduce the F by one-tenth of a dB by shortening the input trace and using a higher Q inductor in L1, but this has not been investigated. 

The prototype is unconditionally stable over 0.05-20 GHz. This means it will not oscillate when either the input or output termination is not 50 ohm impedance outside the target operating frequency - this is an important consideration because aerials and filters are highly reflective outside their passband.  The measured gain (black solid)  exhibited several minor out-of-band “gain peaks” at 10, 13.5 and 18 GHz, but they are well below the unity gain level (0 dB), and so are not expected to potentiate instability. The Rollett stability factor, k (pink trace) is greater than 1 when evaluated up to >20 times the target frequency; meaning that the LNA will be unconditionally stable with any termination having a positive real part.

The prototype demonstrates very high linearity - which helps to minimize spurious reception. Using 1 MHz test spacing, the third order intercept point, OIP3 , exceeds 37 dBm at 900 MHz (fig. 8, red trace). Although the 1296 MHz & 1420 MHz OIP3 were not measured, they are expected to be within a couple of dBs. 

The above data are a subset of the previously published 900 MHz version.