Reproducible and Highly Miniaturized Bazooka RF Balun Using a Printed Capacitor

Abstract

Purpose:

There is currently a strong trend in developing radiofrequency (RF) coils that are high-density, lightweight, and highly flexible. In addition to the resonator structure of the RF coil itself, the balun or cable trap circuit serves as another essential element in the functionality and sensitivity of RF coils. This study explores the development and application of reproducible highly miniaturized baluns in RF coil design.

Methods:

We introduce a novel approach to producing Bazooka baluns with printed coaxial capacitors, enabling the achievement of significant capacitance per unit length. Rigorous electromagnetic simulations and thorough hardware fabrication validate the efficacy of the proposed design across various magnetic field strengths, including 1.5T, 3T, and 7T MRI systems.

Results:

Bench testing reveals that the proposed balun can achieve an acceptable common-mode rejection ratio (CMRR) even when it is highly miniaturized. The use of printed capacitors allows for a notable reduction in balun length and ensures high reproducibility. Findings demonstrate that the proposed balun exhibits no RF field distortion even when placed close to the sample, making it suitable for flexible coils, wearable coils, and high-density coils, particularly in high-field MRI.

Conclusion:

The reproducibility inherent in the manufacturing process of printed coaxial capacitors allows for simple fabrication and ensures consistency in production. These advancements pave the way for the development of flexible coils, wearable coils, and high-density coils.

INTRODUCTION

In the realm of MRI, there is currently a high trend in developing radiofrequency (RF) coils that are high-density, lightweight, and highly flexible^1–12. These advancements are pivotal for enhancing patient comfort, maximizing the signal-to-noise ratio (SNR), and integrating MRI technology into novel applications, such as MRI-guided therapy systems. In addition to the resonator structure of the RF coil itself, the balun or cable trap circuit serves as another essential element in the functionality and sensitivity of RF coils¹³. The primary role of the balun and cable trap is to block undesired common-mode signals present on the coaxial cable’s shield. By doing so, they effectively avoid safety concerns^14–16 and SNR and/or transmit efficiency decrease associated with common-mode currents.

While numerous approaches have been suggested to tackle the hurdles of constructing flexible coils^1–7, achieving complete flexibility remains an ongoing challenge due to the rigidity and bulkiness of the baluns or cable traps linked with coils. This dilemma parallels the scenario where the flag embodies lightness and flexibility, yet the flagpole stands as a representation of bulkiness and rigidity. Most baluns are unsuitable for flexible RF coils due to (1) their large footprint and (2) potential electromagnetic (EM) interactions with the RF field of transmit (Tx) and/or receive (Rx) coils.

In traditional coil setups, RF baluns are typically positioned a few centimeters away from tissue^17–20. However, in flexible coil configurations where the coil and components are placed much closer to tissue, there is a potential for distortion in the RF field around the surrounding human tissues. Although the cable trap can be shielded to prevent magnetic resonance coupling with the RF field, and the Bazooka balun generates a self-constrained EM field that is not likely to couple with the RF field through resonance coupling^20,21, the size of the cable trap or balun circuit remains a crucial factor. If the cable trap or balun circuit is not sufficiently small, the large copper foil used for shielding the cable trap or the outer conductor of the Bazooka balun may change the original RF field.

In previous work, Chai et al.²¹ introduced a method to miniaturize Bazooka baluns and enhance their flexibility by replacing lumped components with a coaxial capacitor made from braids and substituting rigid housing with heat-shrink material. Despite these advancements, we noticed the manual production of coaxial capacitors remains labor-intensive and poses challenges for mass manufacturing. Additionally, the thickness of the heat-shrink tube used for the coaxal capacitor limits the achievable capacitance per unit length, necessitating longer baluns to operate at desired frequencies.

This research introduces a novel approach to constructing highly miniaturized and reproducible Bazooka baluns using a simple and cost-effective manufacturing method. By leveraging a flexible printed circuit board (PCB), we devise a technique for producing coaxial capacitors with significant capacitance per unit length, achieved by employing a dielectric layer as thin as 25 micrometers (μm). This innovation substantially reduces the length of baluns. The precision and reproducibility of PCB manufacturing provide a scalable solution for producing miniaturized Bazooka baluns. Another finding we emphasize is that this design is highly miniaturized and would not distort the RF field, which may occur with traditional cable traps or baluns when placed close to the imaging tissue or sample.

METHODS

Concept

Figure 1A illustrates the various layers of a Bazooka balun using a coaxial capacitor²¹. The coaxial capacitor is inherently dependent on the thickness and dielectric constant of the layer between inner and outer conductors, and the overlapping area of the two conductor layers. Figure 1B illustrates the manually crafting balun using the copper braids as conductors and a heat-shrink tubing as the dielectric layer²¹. This design strategy provides flexibility and reduces weight. Despite diligent efforts to reduce the thickness of the heat-shrink tubing to ~0.5 mm, achieving the necessary capacitance to effectively block common-mode currents required significant lengths for the baluns. Further reduction in the thickness of the heat-shrink tubing in the manually crafting capacitor could be feasible. However, it raises concerns about instability that may lead to shorting between the two conductor layers. Balun lengths of approximately 5 cm, 12 cm, and 20 cm were deemed necessary for 7 T, 3 T, and 1.5 T, respectively²¹. The manual assembly of these baluns necessitates heightened attention to detail and meticulous care, presenting significant challenges in mass production.

Figure 1.

(A): Side-sectional view of Bazooka balun using a coaxial capacitor. (B): Illustration of a Bazooka balun using two copper braids and a heat-shrink tube between them as the coaxial capacitor. (C): Illustration of the proposed design using a double-sided flexible PCB as the coaxial capacitor. (D) The relationship between the capacitance of the printed capacitor and the overlap length ($l_{ol}$) for dielectric layer thickness ($T$) of 500 μm and 25 μm.

To surmount these challenges and further advance the miniaturization of the Bazooka balun, we proposed a design utilizing flexible double-sided PCB as the coaxial capacitor, as shown in Figure 1C. Both the previous flexible capacitor made of copper braids and the proposed printed capacitor are parallel-plate capacitors. Their capacitance can be calculated using the following equation, which ignores the edge effect:

$$C=\frac{{\pi \epsilon }_0{\epsilon }_r\left(D\bullet l_{ol}\right)}{T},$$

where ${\epsilon }_0$ is the permittivity of free space, ${\epsilon }_r$ is the relative permittivity of the dielectric material, $l_{ol}$ is the overlap length, D is the diameter of the balun, and T is the thickness of the dielectric layer, i.e., the distance between the inner and outer copper layer. Figure 1D shows the relationship between capacitance and $l_{ol}$,​ with ${\epsilon }_r$ = 3.4, D= 6 mm, $T$ = 500 μm or 25 μm.

By greatly reducing the thickness of the dielectric layer to only 25 μm, the printed capacitor significantly enhances the capacitance, thereby reducing the total length of the balun. This methodology marks a significant stride toward the development of reproducible and highly miniaturized Bazooka baluns.

EM Simulation

A comprehensive series of full-wave electromagnetic (EM) simulations were conducted using HFSS (Ansys, Canonsburg, Pennsylvania, USA) to offer insights for the actual PCB fabrication process^22,23. Figures 2A, B, and C respectively show the simulation models of the Bazooka balun using printed capacitors for 1.5T, 3T, and 7T. The simulations focus on the effects of varying the total length (l_tot) of the balun as well as variations in the l_ol (i.e., overlap length) between conductors. The l_tot was set to range from 4 to 16 cm at 1.5T, from 2 to 8 cm at 3T, and from 1 to 6 cm at 7T. The purpose of varying the l_ol is to adjust the equivalent coaxial capacitance, enabling the balun to achieve high impedance at the designated frequency. The initial value of l_ol was estimated based on the theoretical analysis of Chai et al.²¹ Depending on different l_tots, the range of l_ol is 0.6–3.6 cm for 1.5T, 0.3–1.8 cm for 3T, and 0.06–0.8 cm for 7T.

Figure 2.

(A)-(C): Simulation models of the Bazooka balun using a printed capacitor at 1.5T, 3T, and 7T. D: Illustration of the cross-sectional view of the balun. The dielectric layer (purple in Figure D) is drawn thicker intentionally for visibility.

As shown in Figure 2D, the baluns have an inner diameter of 6 mm, with their inner and outer conductors separated by a 25μm-thick polyimide film. We selected a diameter of 6 mm for the overall size as it offers an acceptable common-mode suppression performance while also ensuring sufficient minimization in the axial section²¹. We selected these PCB parameters because polyimide is the primary dielectric material for flexible PCBs, and a thickness of 25 μm is nearly the minimum for double-sided flexible PCBs.

Note that a reduction in l_tot results in an increase in l_ol. For example, at 3T, a l_ol that is similar as the l_tot (~1.8 cm) is required if the l_tot is 2 cm. For a Bazooka balun with the specified parameters (diameter 6 mm, using 25-μm-thick polyimide, etc.), minimal l_tots at 1.5/3/7 T are approximately 1/2/4 cm, respectively. In the EM simulation, the common-mode rejection ratio (CMRR) was evaluated as the transmission coefficient (S₂₁) between two 50-ohm ports at the ends of the balun circuit and connected via a ground plane.

Hardware Fabrication and Test

Based on the simulation results (further elaborated in the Results Section), two versions of such baluns were fabricated, as illustrated in Figure 3. One version aims to achieve a CMRR better than −15 dB, with l_tot of 8cm/4cm/2cm for 1.5T/3T/7T, respectively. The other version was designed with almost minimal length, featuring l_tot of 4cm/2cm/1cm for 1.5T/3T/7T, respectively.

Figure 3.

(A) Cross-sectional view of a flexible PCB showing the top and bottom conductor layers with the dielectric film forming a capacitor. (B)-(D) Drawings of double-sided PCBs with different lengths (8/4/2 cm long version and 4/2/1 cm long version) designed for different static magnetic fields (7T, 3T and 1.5T). The dark brown color represents the copper layers with coverlays, while the light brown color indicates the pads for soldering. (E) Photographs of fabricated baluns.

The completed flexible PCBs have a thickness of 0.11 mm, comprising two layers of conductors, each 12 μm thick, and a dielectric layer of 25 μm, with the remaining portion consisting of adhesive and insulation layers. The flexible PCB was coiled around a 3D-printed hollow cylindrical support (Forms3, Formlabs), as shown in Figure 3. These PCBs feature extended pads at both ends, designed to facilitate soldering the PCB to the cable’s shield. All baluns were fabricated on a non-magnetic ultra-flexible cable (Huber + Suhner G_02232_D), which shares similar dimensions to RG-174.

The workbench test was conducted using a calibrated Vector Network Analyzer (VNA) (E5071C, Keysight, Santa Rosa, CA). CMRR was directly measured by attaching each end of the cable’s shield to a respective port on the VNA, a method detailed in references^19,24. The CMRR was subsequently evaluated based on the S₂₁ measurement between the two VNA ports, with the VNA being calibrated (response calibration) such that an S₂₁ value of 0 dB corresponded to scenarios where the balun was absent.

To demonstrate the reproducibility of this design, we fabricated and tested eight 1-cm long 7T baluns using the same procedures to assess their consistency. First, we applied double-sided tape to the back of the flexible PCB and wrapped it around the 3D-printed cylinder. The two ends of the PCB were isolated by the insulator layer and not soldered together. Kapton tape was used near the slit between the two ends of the PCB to ensure it adhered well to the cylinder. Afterward, we bent the extended pads and soldered them to the cable shield. If needed, the extended pads could be slightly adjusted to finely tune the operating frequency to the desired one. Finally, the balun was sealed with heat-shrink tubes to isolate it from the outside. Step-by-step fabrication procedures are shown in Supporting Figure S1.

To further illustrate the proposed balun, we fabricated 7T coils with and without a 1cm-long balun using the printed capacitor, as shown in Figure 4. We measured the impedance changes when touched by hand. The coils, sized 10 × 10 cm², were tuned to 298 MHz (the Larmor frequency at 7T) and well-matched to 50 ohms, with return loss (S₁₁) < −30 dB. The changes in S₁₁ were recorded when the cable’s outer jacket was touched by hand.

Figure 4.

Circuit diagram and photograph of 7T coils without and with the proposed balun.

As mentioned above, if positioned too close to the tissue, another concern regarding the balun is its potential to induce changes in the RF field. To comparably investigate such effect, we conducted GRE scans on a 12-cm-diameter bottle-shaped oil phantom at 7T, without any balun, with a relatively large shielded cable trap (15 mm diameter and 50 mm long, referred to as Trap_D15L50), with a relatively small shielded cable trap (11 mm diameter and 24 mm long, referred to as Trap_D11L24), with a small float Bazooka balun (12 mm diameter and 25 mm long, referred to as Float_D12L25) and with the highly- minimized 1-cm-long balun (referred to as Print_D06L10). Two kinds of coils were investigated: a volume Tx/Rx coil and an 8Tx/32Rx coil (both from Nova Medical Inc., Wilmington, MA) with default fixed RF weights. In all scenarios, the balun was placed directly above the phantom. The sizes of cable traps or baluns were chosen to be similar to or smaller than those widely used in previous publications^24–28. The parameters of the GRE scan are: axial slices, TR/TE = 100/2.4 ms, flip angle = 90/30 degrees for Nova volume/8Tx coil, field of view = 150 × 150 mm², matrix = 152 × 152, slice thickness = 2 mm, slice gap = 6 mm, number of averages = 1.

RESULTS

Figures 5A–C illustrate the correlation between the CMRR and l_tot in simulations. As expected, CMRR improves with increasing l_tot. To achieve a CMRR of −15 dB or better (indicating >97% common-mode signal attenuation), l_tot should be at least 8 cm, 4 cm, and 2 cm for 1.5T, 3T, and 7T, respectively, as depicted in Figures 5D–F. However, it is noteworthy that even with the minimum l_tot values of 4/2/1 cm for 1.5T/3T/7T, the simulated CMRR could reach −10 dB or better, indicating more than 90% of common-mode signal attenuation, as shown in Figures 5G–I.

Figure 5.

(A)-(C) Simulated CMRRs versus the balun’s total length (l_tot) at different static magnetic fields. (D)-(F) Simulated CMRR plots with l_tot of 8/4/2 cm for 1.5/3/7 T. (G)-(I) Simulated CMRR plots with l_tot of 4/2/1 cm for 1.5/3/7 T.

Figures 6A–C display the measured CMRR for baluns with l_tot of 8/4/2 cm, while Figures 6D–F display the measured CMRR for baluns with a minimum l_tot of 4/2/1 cm. For the 8/4/2 cm versions, the measured CMRRs are −19.7, −16.7, and −15.9 dB at 64 MHz, 128 MHz, and 298 MHz, respectively. For the 4/2/1 cm version, the measured CMRR can still achieve −13.2, −15.3, and −14 dB at 64 MHz, 128 MHz, and 298 MHz, respectively. These results validate the effectiveness of the proposed balun, demonstrating their capability to effectively suppress common-mode signals across various frequencies within a highly miniaturized structure. Overall, the bench test results are consistent with the simulation results and further validate the concept.

Figure 6.

(A)-(C) Measured CMRRs of baluns with lengths of 8/4/2 cm for 1.5/3/7 T. (D)-(F) Measured CMRRs of baluns with lengths of 4/2/1 cm for 1.5/3/7 T.

Figure 7 summarizes the measured CMRRs and operating frequencies of all eight baluns fabricated under the same conditions. These baluns exhibit consistent operating frequencies (with less than 0.5% deviation from 298 MHz) and CMRR values (ranging between −12 dB and −14 dB). The CMRR plots versus frequency for all these baluns are shown in Supporting Figure S2.

Figure 7.

The measured CMRRs (A) and operating frequencies (B) of all eight baluns fabricated under the same conditions.

Figure 8 illustrates the S₁₁ changes due to the “hand effect” for coils at 7T. The coil was first well-tuned to 298 MHz and matched to 50 Ω, with an S₁₁ of −37 dB (negligible return loss). Without a balun circuit, notable impedance change was observed when the cable was touched by hand (also known as the ‘hand effect’). This impedance change could be attributed to the common-mode current forming a secondary ground loop when the balun is absent. This effect resulted in impedance shifts from −37 dB to −14 dB, equating to a return loss of 4%. However, with the implementation of printed baluns, the coils’ impedance was maintained at −29 dB (equivalent to ~0.1% return loss), further demonstrating the printed balun’s effectiveness in reducing undesirable common-mode currents.

Figure 8.

Measured impedance matching (S₁₁) change of the 7T coils caused by touching the cable’s outer jacket with a hand. A: S₁₁ baseline. In this case, the hand does not touch the cable. B: S₁₁ plots of the coil with no balun circuit in the presence of a hand. C: S₁₁ plots of the coils with the proposed balun in the presence of the hand.

Figures 9A and 9B depict the measured single-slice GRE images acquired from the oil phantom at 7T using the Nova birdcage volume coil and the Nova 8Tx/32Rx coil, respectively. The multi-slice GRE images are shown in Supporting Figure S3. In both scenarios, the Trapped_D15L50 balun, notable for its larger dimensions, exhibited significant distortions in the area near the balun. Conversely, the Print_D06L10 balun, featuring a highly miniaturized printed design, demonstrated no imaging artifacts under either coil configuration.

Figure 9.

Measured GRE images by using the Nova birdcage volume coil (A) and the Nova 8ch Tx coil (B) without any balun, with a large cable trap (Trap_D15L50), a small cable trap (Trap_D11L24), a small float balun (Float_D12L25) and the proposed highly- miniaturized balun (Print_D06L10). White spots within the images indicate the position of different baluns relative to the phantom.

For Trap_D11L24 and Float_D12L25, whose sizes are between Trapped_D15L50 and Print_D06L10, they did not impair the images of the Nova volume coil but slightly affected the images of the Nova 8ch coil. While the impact of these smaller baluns on GRE images was less pronounced compared to the larger balun, changes in RF fields suggest potential modifications to the electric field, raising safety concerns, particularly in proximity to tissues.

DISCUSSIONS

One major innovation presented in this work lies in its easily reproducible and cost-effective design. The average cost of such a flexible balun amounts to less than ten US dollars, covering expenses for both the PCB and 3D printed support.

Another notable innovation is the achievement of high miniaturization. By utilizing an ultra-thin layer within the printed capacitors (25 μm), a substantial capacitance of tens of pF per centimeter is attained. Consequently, this permits a significant reduction in the total length of the balun. For instance, whereas previous designs for a 3T application demanded a balun length of up to 12 cm, integration of printed capacitors allows for a reduction to a mere 2 cm. It is important to highlight that despite this considerable reduction in size, both designs demonstrate a comparable level of CMRR. Specifically, the previous 12-cm-long design utilizing copper braids achieved a CMRR of −16.4 dB, while the highly miniaturized 2-cm-long design employing flexible PCB attained −15.3 dB. While this design sacrifices some flexibility inherent to the balun itself, its highly miniaturized form (only 1 cm long for 7T) allows it to maintain overall cable flexibility in high fields. Note that if the balun must provide ultra flexibility across every part of the cable, this printed balun does not meet the requirements, and one can still resort to utilizing the previous design outlined in Chai et al.²¹

The highly miniature nature of the balun conserves valuable space and reduces bulkiness in flexible and/or wearable coils. Beyond space-saving benefits, another advantage observed with miniature baluns is their minimal impact on the RF field, which proves particularly crucial in high-field environments. In high MRI systems, identical physical dimensions correspond to significantly larger electrical sizes. For instance, while a 5 cm dimension is approximately 1/100 of the wavelength at 64 MHz (free space wavelength 4.7 meters), it could signify 1/20 wavelength at 300 MHz (free space wavelength 1 meter). This increased electrical size amplifies the likelihood of the balun affecting the Tx field, potentially leading to RF distortion when positioned close to the imaging tissue. Given the unavoidable proximity to the coil in flexible and/or wearable coils, the significance of minimizing the balun’s impact is further accentuated.

Incorporating a balun or cable trap circuit immediately after the coil is the optimal method for suppressing common mode current, as any conductor (including the cable) preceding the balun or cable trap could be perceived as part of the coil. By meticulously designing the coil circuit and thoughtfully arranging the cable along the virtual ground plane, it may be feasible to position the cable trap or balun remotely. However, this approach typically requires ensuring the high symmetry and balance of the loop and feeding circuit, as well as meticulous cable routing.

If the balun is not properly constructed, it may significantly increase the temperature near the balun and cable, potentially causing safety concerns^14–16. Therefore, in this study, we conducted a temperature test on our proposed balun in a 7T scanner under a high SAR sequence. We found that our balun did not exhibit any temperature rise. The details of the test and the results are shown in Supporting Figure S4 of the supplementary information. (C 2.2) The breakdown voltage of polyimide (PI) film is greater than 500 V/μm^29,30. In our work, the thickness of the PI film used in the PCB is 25 μm, resulting in a breakdown voltage that can reach 10,000 V. This breakdown voltage is significantly above the voltage levels expected in the balun circuit. Therefore, we believe the balun is unlikely to experience breakdown or arcing. Our 7T experiment also confirmed this, as we observed no breakdown or arcing. However, we acknowledge that the breakdown voltage also depends on the design of the top and bottom traces. In our design, the top and bottom traces are approximately a couple of centimeters wide, which is unlikely of generating arcing. The only potential concern might arise if the two ends of the board (wrapped around the cylinder) are too close to each other (<10 μm), which could create a high electric field between the edges and potentially lead to a breakdown. To ensure safety, we suggest maintaining a gap of 0.5 mm or greater between the two ends of the PCB. (C 2.1)

Some limitations within this study need to be mentioned. Firstly, the non-float design of the highly miniaturized balun presents practical challenges concerning its adjustability post-fabrication onto the cable. This limitation stems from the necessity of soldering or clamping the balun onto the cable shield. Another limitation is that the miniaturized design makes it inadequate for accommodating the thick multi-core cables commonly associated with ODU plugs in MRI. To effectively suppress common-mode current along multi-core cables, alternative solutions such as the standard float Bazooka balun²⁰ or the recently proposed ‘caterpillar’ float balun³¹.

In this study, we utilized an oil phantom to investigate RF artifacts, aiming to circumvent the intrinsic ‘Tx artifacts’ present in water due to destructive interfaces. Beyond the dimensions of the balun itself, various factors may contribute to RF distortion, including the distance between the balun and the loading, the structure and size of the transmit or receive coil, and the separation between the balun and the coil. Further research is needed to explore how balun size influences the RF field across different transmit coils.

CONCLUSION:

This paper introduces a novel balun, miniaturized and easily replicable, fabricated using printed coaxial capacitors. Through simulation and bench testing, it has been confirmed that despite its significantly reduced size, this balun maintains a commendable CMRR. Such a development lays a solid foundation for the creation of high-density, lightweight, and flexible coils.

DATA AVAILABILITY STATEMENT

Complete resources required to build such highly miniaturized bazooka RF balun using a printed capacitor are freely available at https://github.com/XinqiangYan/Printed-Bazooka-Balun. Free samples of such baluns could be obtained from the corresponding author upon reasonable request.