Wideband pre-amplifier using discrete transistor

Introduction

Wideband preamplifiers, especially hobbyists', are most commonly realized with microwave monolithic integrated circuits (MMICs) [1]-[2]. But MMICs capable of operating down to LF/HF, such as the Mini-circuits MAR-6 / Avago MSA-06xx or ABA-52563 [3], have noise figures (NF) above 3 dB (link to the MAR-6 / MSA-06 wideband pre-amp). GaAs pseudomorphic high electron mobility transistor (PHEMT) MMICs, such as the MGA-82563 and the MGA-30889, offer lower NF of ~2dB, but cannot operate below 40-50MHz [4]. A lower noise figure can improve reception in the VHF-UHF range where atmospheric noise is low. To secure a lower NF from HF to UHF, a pre-amplifier using a discrete silicon bipolar junction transistor (BJT) is constructed and characterized. This article documents the pre-amplifier implementation and the achieved performances. The design is far from optimum, but it is documented here so that its weaknesses along with possible fixes can be made known to others.

Materials & methods

To achieve a wide bandwidth in the preamplifier, a BJT with 10GHz ft is selected. The AT-32033 has 20 emitter fingers and a 3.2 micron emitter-to-emitter pitch [5].  However, the choice of BJT is not critical due to the application of feedback - hence, the AT-32033 can be substituted with other low noise BJTs, as shown below (fig. 7)

The AT-32033's intended application is tuned, narrowband LNA at UHF & lower microwave. To flatten the gain for wideband operation and to force the input/output impedances to 50R, resistive shunt feedback is applied (fig. 1). On the flip side, lossy feedback degrades NF and output power at 1dB gain compression (P1dB). Since masthead preamplifiers are commonly supplied with 12V supply, this design's collector resistors are dimensioned to drop the 12V supply down to the target Vce. Ferrite beads are slipped over each of the 180R resistor's leads to improve the choking at higher frequencies - the intention is to increase the upper frequency limit.

The device operating point is chosen as a compromise between noise, large signal performance and tolerance of voltage fluctuation. The datasheet characterizes the NF and s-parameters at different Icc in the 1 to 20 mA range. Although dynamic range improves with bias current, the second highest value, 10mA, is chosen because the maximum value, 20mA, is accompanied by a ~0.4dB jump in NF (fig. 2). The datasheet includes characterization at 2.7V and 5.0V Vce. The 5.0V Vce data shows a 2 dB higher P1dB, but it is so close to the device's max. Vce of 5.5V, that a regulated supply will be necessary. So, Vcc= 2.7V is chosen because the wider margin allows powering from an unregulated wall wart.

The model consists of the transistor, passive components and PCB traces (microstrip). However, the BNC connectors were not modeled because they were not expected to significantly impact the RF performances.

The feedback resistance (R2) was chosen to optimize matching at both 30 MHz (fig. 4) and 435 MHz (fig. 5). At 30 MHz, simulation predicts that the best matching occurs over 480-700 ohm. Whereas, at 435 MHz, matching is best over 400-700 ohm. So, as a compromise, a value of 560 ohm was chosen. 

Aside from matching, the feedback resistance also determines the 3 dB bandwidth (BW). The chosen 560 ohm resistance is predicted to result in a ~430 MHz BW (fig. 6).  

Based on the aforementioned 560 ohm feedback resistance, the performances of low noise BJTs from other manufacturers  were simulated (fig. 7). These BJTs can replace the AT-32033 without substantially changing the frequency response.

To enclose and to RF-proof the amplifier, a 50x30x20 mm box is fabricated from thin brass sheet (fig. 8). RF connections are made through BNC connectors. To filter conducted noise on the power supply, a feedthru capacitor mounted on the case is used as the positive supply terminal.

Results & discussion

Both input and output are well matched over several octaves. Both gain and return losses degrade abruptly below 20MHz due to the small value capacitors used. With hindsight, a larger value like 100nF should have been chosen as it would have beneficially extended the lower frequency limit down to long-waves. The simulated input return loss agrees with the experimental with less than 3 dB error over 0 - 500 MHz. However, the simulated output return loss has significant error below 400 MHz - this probably due to the transistor model being created from s-parameters that start above 0.5 GHz (refer to datasheet's s-parameter tables).

The experimental NF measures less than 1.7dB over the evaluated range while the simulated result is ~0.4 dB higher.  Trading off bandwidth, the achieved result is ~1 dB better than Mini-circuits MAR6 / Avago MSA-0686 which uses a similar device technology.

The experimental gain roll-off by 3 dB at 400 MHz. Therefore, the design's usable bandwidth (BW) is limited by its gain. Compared to the MAR6/MSA-0686 amplifier, the BW is ~50% less but the min. gain is ~2dB higher. The simulated gain is ~ 2 dB lower, but agrees qualitatively with the experimental.

Between 250MHz and 500MHz, the gain slope is 2.5dB/octave - the less than 6dB/octave gradient is likely due the 'shunt peaking' effect of the ferrite beads.

The prototype is unconditionally stable over 0-6 GHz. The measurement shows that both k>1 and B>0 stability criteria are satisfied. The simulated stability agrees well with the experimental.

The experimental P1dB is greater than 1.5 dBm over a 10-300 MHz range (fig. 12). This is fractionally lower than the MAR6/MSA-0686’s. The modeled P1dB shows the same trend but is ~ 2 dB lower. The AT-32033 is capable of 12dBm P1dB in narrowband mode, but in making it broadband, substantial output power is lost in the feedback and the resistive drain load. This humble blocking performance is only adequate for small aerials and when there is no nearby VHF broadcast (TV/FM) station or ham neighbour.

The third order output intercept point (OIP3) is not tested but it is typically assumed to be ~10dB higher than the P1dB. The simulated OIP3 is 11 dBm over 10~500 MHz (fig. 13). Concerns arising from the modest P1dB (and OIP3) are blocking and intermodulation. So, the output spectrum at the aerial terminal should be evaluated first to ensure that largest peaks are below the pre-amp's limit. If strong signals are received at the installation, a higher IP3 design, such as the Norton or Griffith feedback, should be considered instead. 

Conclusion

A discrete implementation of the wideband pre-amplifier can offer less noise compared to the more popular silicon MMIC implementations. Because the amount of negative feedback can be adjusted externally, the constructor can trade off bandwidth for noise figure. Ideas worth investigating in the next iteration are: larger capacitances to remedy the current prototype’s low frequency roll-off, and phantom powering via the RF output. We anticipate that discrete designs, combined with better device technology, will lead to lower noise in wideband receive preamplifiers.

References

[1] OM6NU, "RE-HFA1MAR6 Wideband VHF/UHF/SHF monolithic PreAmp based on MARx-series," [Online] Available: http://users.belgacom.net/hamradio/schemas

[2] A. Palmer, "Improved VHF-UHF Masthead Amplifier," Electronics Australia, Jun. 1998.

[3] Avago Technologies product specification, “ABA-52563 3.5 GHz Broadband Silicon RFIC Amplifier,” [Online] Available: http://www.avagotech.com

[4] Avago Technologies product specification, “MGA-82563 0.1– 6 GHz 3V, 17 dBm Amplifier,” [Online] Available: http://www.avagotech.com

[5] Avago Technologies product specification, "AT-32011, AT-32033 Low Current, High Performance NPN Silicon Bipolar Transistor," [Online] Available: http://www.avagotech.com

(c) This work is protected by copyright and may not be reproduced in any form without the expressed consent of the owner.