Simple 10-2000 MHz biastee

Created: Jul. 2012. Updated: May 2024.

Introduction

Wideband biastees have two major areas of application: (i) RF semiconductors characterization fixture and (ii) phantom powering of masthead pre-amplifier in wireless reception. The aim is to build a biastee covering 10-2000MHz using components in the junk box.

Material and method

A microstrip transmission line bridges the 'RF' and 'RF & DC' ports to form the biastee's signal path (fig 1). For economy, the microstrip consists of a 1.5 mm thick double-sided FR4 PCB. To minimize reflection at the coax-microstrip transitions, the trace width is scaled for a 50 ohm characteristic impedance. To minimize insertion loss, the two RF ports should be positioned as closely as possible because microstrip loss scales linearly with length - hence, it is preferable to house the biastee in a small enclosure (fig. 2). The 10nF DC blocking chip capacitor (C1), a 4.6 uH ferrite ring inductor (L1) and a ferrite bead (L2) are mounted on the microstrip side of the PCB. The ring core is not the optimum shape for a wideband choke - commercial biastees use conical solenoids for the highest Self Resonance Frequency [1] - but the ring is readily available in my junk box :-). For the widest bandwidth, a small core with high permeability will be best - however, a moderate permeability core (ui=125) was used because it is available; i.e. the toroid material & size are not critical. A ferrite core enables a higher self resonance frequency than an air core because inter-winding capacitance is reduced by the former's fewer turns and smaller size. The flip side of a small ferrite core is reduced current carrying capability because of core saturation - however, this doesn't pose a problem for phantom powering a pre-amp because of the small current. 

The ferrite bead L2 was used in the hope of improving the high frequency response beyond L1's lowish SRF. However, later circuit simulation shows that the bead's role is ambiguous and so, it can be removed without detriment.

To improve RF-DC isolation, a 1nF feed-through capacitor C2, mounted in a PCB hole and soldered to the ground plane, decouples the 'cold' end of the RF choke. Therefore, the PCB ground plane also functions as an RF shield to improve the isolation between RF path and the DC port. A second, non-critical, ferrite ring inductor L3, which connects the feed-through capacitor to the 'DC' port, improves the RF-DC isolation at low frequencies. A Zobel network (R1 & C3) consisting of 56R and 22nF in series, improves the DC port's high frequency match. 

The biastee enclosure is fabricated from single-sided FR2 PCB. The seams are soldered. The size is 3 x 4 x 2 cm (fig 2). 

An equivalent circuit model was created to simulate the RF performances (fig. 3). The simulation also reveals how component values affect RF performances.  In the equivalent circuit, components  SRL1 and SRL2  are used to represent the BNC connectors' parasitic inductance, Lbnc. TL1-2 represents the microstrip traces on the PCB.  The combination of L2a and L2b form the ferrite bead L2. For convenience, C2, C3, L3 and R1 are omitted from the model. 

Simulation reveals that the BNC connectors' parasitic inductances greatly influence performances. Higher stray inductances will worsen the matching (RL). The degradation is most pronounced at higher frequencies while frequencies <50 MHz are unaffected (fig. 4). Likewise, insertion loss (IL) at high frequencies is worsened by larger Lbnc values (fig 5).  Lbnc = 2 nH provides the best fit to measurement. 

The biastee's lower frequency limit is imposed by L1.  A larger L1 value will extend the minimum frequency (fig. 6).  The 4.6 uH chosen value is sufficient for operation down to 10 MHz. 

Results

The prototype's IL is better than -1 dB over 10-2000 MHz (fig. 7). The simulated IL (red trace) shows a similar trend to the measurement (black trace). The simulation's error is < 0.1 dB. 

The prototype's RL is better than -12dB over 10-2000MHz (fig. 8). Both 'RF' and 'RF & DC' ports exhibit almost similar RL. Simulation predicts an RL minima that is ~30 MHz lower than actual.  Otherwise, the simulated result has the same trend as the experimental. The simulated result’s error is ≤ 6dB.

The prototype demonstrates greater than 40dB isolation between RF & DC ports. Peak isolation occurs around 200MHz - this coincides with the the ferrite ring inductor's self resonance. At the lower frequency limit, the ferrite ring inductor's decreasing reactance degrades the isolation. Above 200MHz, the slowly degrading isolation is probably caused by insufficient choking and RF shielding.

Conclusion

A relatively simple biastee can be designed for good performance over 10-2000 MHz. The RF performances can be predicted quite well using simulation. 

Reference

[1] T. A. Winslow, "Conical inductors for broadband applications," IEEE Microwave Magazine,  Mar. 2005.