
AiP vs AoB: Which mmWave Antenna Architecture Fits Your Design?
Antenna-in-Package (AiP) places the radiating elements—patches, slots, or dipoles—inside or on the surface of the IC package substrate, typically using low-loss organic laminates or low-temperature co-fired ceramic (LTCC). Antenna-on-Board (AoB) etches those same elements directly onto the system-level PCB, using high-frequency laminates such as Rogers RO3003 (Dk 3.00, Df 0.0010 at 10 GHz) or Isola Astra MT77. The choice between AiP and AoB affects gain, loss budget, thermal management, module size, cost structure, and time to market—and the right answer depends on frequency, volume, and system constraints.
How AiP Works

In an AiP module, the RFIC die is flip-chip bonded or wire-bonded to a multi-layer package substrate. Antenna elements are formed on the top metal layer or embedded within the substrate stackup, and the ground plane sits one or two dielectric layers below. The entire assembly is often covered by a molding compound or a dielectric superstrate that also serves as a radome. Solder balls on the bottom connect the module to the system board.
Because the interconnect between the RFIC output and the antenna feed is measured in hundreds of micrometers rather than millimeters, AiP dramatically reduces feed-line loss. At 60 GHz, a 10 mm microstrip run on FR-4 can lose over 2 dB; inside a package on a low-loss BT-resin or ABF substrate, the same electrical path might be under 0.3 dB. That saved loss translates directly into higher EIRP or, equivalently, lower TX power for the same link budget.
Typical AiP Materials and Stackups
- Organic substrates (BT resin, ABF): Dk 3.2–3.5, Df 0.005–0.008 at 28 GHz. Cost-effective for high-volume fan-out packaging.
- LTCC (e.g., Ferro A6M): Dk ~5.9, Df ~0.002 at 60 GHz. Excellent dimensional stability and hermeticity, but higher per-unit cost and longer lead times.
- HDI organic with embedded cavities: Allows air-cavity antenna elements for wider bandwidth, used in some 5G FR2 modules.
Texas Instruments’ AWR series 77 GHz radar transceivers are a canonical AiP example: the AWR2944 integrates a 4-TX, 4-RX radar front end with the antenna array built into the package, simplifying the radar front-end architecture considerably.
How AoB Works
In an AoB design, the RFIC is packaged in a conventional BGA or QFN without any radiating elements. Antenna arrays—patch, slot, or more exotic topologies—are lithographically defined on the system PCB. The RFIC output routes through a controlled-impedance transmission line (microstrip, stripline, or substrate-integrated waveguide) to the antenna feed network, all on the same board.
AoB gives the antenna designer full control over the PCB laminate, stackup, and array geometry. At 28 GHz, a 4 × 4 patch array on Rogers RO4835T (Dk 3.48, Df 0.0037 at 10 GHz) can achieve 18–20 dBi directivity with beam-steering via phase-shifted feeds. The trade-off is that the feed network between RFIC and array may span 15–30 mm, and every millimeter of line adds loss—roughly 0.15–0.25 dB/mm on microstrip at 28 GHz on RO4835T, depending on trace width and copper roughness.
Common AoB Laminate Choices
| Laminate | Dk (10 GHz) | Df (10 GHz) | Typical Use |
|---|---|---|---|
| Rogers RO3003 | 3.00 | 0.0010 | 77 GHz automotive radar |
| Rogers RO4835T | 3.48 | 0.0037 | 28/39 GHz 5G arrays |
| Isola Astra MT77 | 3.00 | 0.0017 | mmWave backhaul, V-band |
| Panasonic Megtron 7 | 3.37 | 0.0020 | High-speed digital + RF hybrid |
The flexibility of AoB also enables advanced radiator types. A SIW slot radiator at 60 GHz, for example, can be routed as a substrate-integrated waveguide directly in the PCB copper layers, avoiding the radiation leakage of open microstrip feeds. Similarly, waveguide-fed antenna transitions become practical when the board real estate is available.
Head-to-Head Comparison: AiP vs AoB
| Parameter | AiP | AoB |
|---|---|---|
| Feed-line loss (28 GHz, typical) | 0.2–0.5 dB | 1.5–4.0 dB |
| Antenna bandwidth | Moderate (constrained by package Dk and thickness) | Wide (designer picks laminate and stackup) |
| Array size flexibility | Limited by package outline (often ≤ 12 × 12 mm) | Limited only by board area |
| Thermal path | Short to heatsink via package lid or thermal vias | RFIC heat must traverse BGA + board |
| Module size | Very compact; antenna + RFIC in one BGA | Larger; RFIC + feed network + array on PCB |
| NRE cost | High (custom package design, mold tooling) | Low (standard PCB fabrication) |
| Per-unit cost at volume | Lower at > 100 k units (amortized NRE) | Lower at < 10 k units |
| Design iteration speed | Slow (package re-spin weeks to months) | Fast (board re-spin days to weeks) |
| Repairability | Module-level replacement only | RFIC or antenna section can be reworked |
Loss Budget: Where the Decibels Go
Understanding the mmWave antenna loss mechanisms is essential for choosing between AiP and AoB. In an AoB system at 77 GHz, the loss budget typically breaks down as follows:
- BGA ball and pad transition: 0.3–0.6 dB
- Microstrip or GCPW feed line (20 mm): 3–5 dB
- Power divider / corporate feed network: 0.5–1.5 dB
- Patch element mismatch and surface-wave loss: 0.3–0.8 dB
Total: roughly 4.5–8 dB from RFIC output to radiated power. An AiP module at the same frequency compresses the feed line to under 1 mm and eliminates the BGA transition entirely, cutting total feed loss to 0.5–1.5 dB. That 3–6 dB advantage is why nearly all commercial 60 GHz WiGig and 77 GHz automotive radar modules use AiP.
When to Choose AiP
- Frequency above 60 GHz: Feed-line losses on PCB substrates become prohibitive. At V-band (40–75 GHz) and above, AiP is almost mandatory for competitive EIRP.
- High-volume consumer or automotive: Amortized NRE makes AiP cheaper per unit when volumes exceed 100 k. Automotive 77 GHz corner radar is a textbook case.
- Size-constrained products: Smartphones, wearables, and IoT sensors where a 10 × 10 mm module footprint is the entire RF section.
- Standardized radio interfaces: When the antenna pattern and gain are well-defined by the standard (e.g., IEEE 802.11ad/ay), a pre-qualified AiP module accelerates certification.
When to Choose AoB
- Frequency at or below 28–39 GHz: PCB feed losses are manageable, and the designer gains full control over array aperture. 39 GHz 5G n260 boards routinely use AoB with 64- or 128-element arrays.
- Large aperture / high gain needed: A 256-element array for fixed wireless access at 28 GHz simply will not fit inside a package. AoB is the only practical path.
- Prototype and low-to-medium volume: Board-level iteration is faster and cheaper. A new antenna layout can go from HFSS simulation export to fabricated board in under two weeks.
- Multi-function boards: When the antenna shares a board with baseband, power management, and connectors—common in small-cell 5G RAN radio units—AoB avoids the added connector or flex interface an AiP module would need.
- Custom beam patterns: Shaped beams, null steering, or non-uniform array spacings are straightforward to implement on PCB; re-spinning a package for each variant is cost-prohibitive.
Hybrid Approaches: AiP + AoB on the Same System

Many real-world mmWave products blend both architectures. A common pattern in 5G FR2 base stations is an AiP module containing the transceiver and a small 2 × 2 or 4 × 4 sub-array, soldered onto a larger motherboard that carries a board-level feed network combining multiple AiP modules into a full 256-element panel. This “AiP tile on AoB combiner” approach captures the low feed loss of AiP for the last-centimeter connection while exploiting AoB’s scalability for the large-aperture combining network.
Another hybrid is the “bare-die on board” method, where the RFIC die is flip-chip mounted directly onto the antenna PCB, eliminating the package entirely. This achieves AiP-like feed-line lengths with AoB’s laminate freedom, but demands gold-stud or copper-pillar bumping and underfill—processes that push PCB assembly into semiconductor-grade territory.
Thermal Considerations
AiP modules concentrate heat in a small volume. A 77 GHz radar RFIC dissipating 2–3 W inside a 12 × 12 mm BGA needs a well-designed thermal via array and often a heat spreader lid. The advantage is a short, low-resistance thermal path to the heatsink. In AoB, the same RFIC sits in its own package with dedicated thermal pads, and the antenna area of the board does not generate significant heat—but the RFIC’s thermal path now includes the BGA solder joint plus the system board’s thermal vias, adding thermal resistance.
For designs where the antenna and RFIC share a thermally constrained enclosure—automotive radar behind a bumper fascia, for instance—AiP’s compact thermal path is a clear advantage. For base-station panels with forced-air cooling and generous board area, AoB’s distributed layout keeps junction temperatures manageable without exotic packaging.
Manufacturing and Test Implications
AiP Manufacturing
AiP production involves wafer-level or panel-level packaging, redistribution layers (RDL), and often embedded component integration. Yield is measured at the package level, and a failed antenna element means scrapping the entire module. Over-the-air (OTA) testing at the package level requires a compact antenna test range or near-field probe system calibrated to the operating frequency—capital equipment that adds six to seven figures to the test line.
AoB Manufacturing
AoB leverages standard PCB fabrication: lamination, drilling, plating, etching. Antenna elements are just copper features, inspectable with automated optical inspection (AOI). RF performance is verified after assembly by antenna pattern measurement at the board or system level. The lower barrier to entry makes AoB accessible to any design house with mmWave PCB fabrication capability, whereas AiP typically requires partnership with an OSAT (outsourced semiconductor assembly and test) provider.
IPC standards apply differently too. AoB boards follow IPC-6012 for rigid board qualification, with Class 3 or 3/A for high-reliability applications. AiP substrates fall under semiconductor packaging standards like JEDEC and IPC/JEDEC J-STD-020 for moisture sensitivity, adding another layer of qualification.
Cost Structure Summary
| Cost Element | AiP | AoB |
|---|---|---|
| NRE (package design, tooling) | $50 k–$500 k+ | $5 k–$30 k |
| Laminate / substrate material | Included in package cost | $30–$150 per panel (high-freq laminate) |
| Assembly complexity | High (flip-chip, underfill, mold) | Standard SMT reflow |
| Test capital | OTA probe at package level ($$$$) | Board-level VNA + chamber ($$$) |
| Crossover volume (AiP cheaper) | Typically 50 k–200 k units, depending on frequency and array size | |
Design Decision Flowchart
- Is the operating frequency above 60 GHz? → Strongly favor AiP. PCB feed losses are too high for competitive performance.
- Is the array aperture larger than ~15 × 15 mm? → Favor AoB. Package sizes above this are exotic and expensive.
- Is production volume above 100 k units/year? → AiP NRE amortizes well; evaluate total cost of ownership.
- Do you need rapid design iteration? → Start with AoB for prototyping; migrate to AiP for production if volume justifies it.
- Is the product a multi-function board with digital + RF? → AoB avoids an extra module-to-board interface.
Frequently Asked Questions
Can AiP and AoB achieve the same antenna gain?
For a given array size, yes—the radiating physics are identical. The difference is in feed-network loss. AiP delivers more of the generated power to the aperture, so for the same EIRP target, an AoB array may need a slightly larger aperture or higher TX power to compensate for 2–5 dB of additional feed loss.
Is AiP always better for 5G mmWave phones?
In practice, yes. Smartphone 5G FR2 modules from Qualcomm (QTM545/547) use AiP because the antenna must fit within a 10 × 5 mm module outline and operate at 28 or 39 GHz. AoB would require dedicating mainboard area and a high-frequency laminate zone, which conflicts with the dense, multi-layer HDI stackups phones use.
What is the maximum practical frequency for AoB?
AoB has been demonstrated up to 77 GHz on RO3003 and similar ultra-low-loss PTFE laminates, but feed-line lengths must be kept under about 5 mm to stay within a reasonable loss budget. Above 100 GHz, AoB is largely impractical without bare-die-on-board techniques.
Does AiP eliminate the need for a radome?
Not entirely. The mold compound or package lid acts as a near-field dielectric layer, and its Dk and thickness must be carefully controlled—much like antenna radome Dk and thickness budgeting on a board-level design. Some AiP modules add an external radome for environmental protection, requiring the same electromagnetic analysis.
Can I prototype in AoB and then move to AiP for production?
This is a common and recommended strategy. Validate the array pattern, beamforming weights, and link budget on an AoB prototype where iteration is cheap, then port the proven antenna geometry into a package substrate for volume production. Expect one to two re-optimization cycles because the package dielectric environment differs from the PCB laminate.
Ready for a quotation?
Send your requirements for a material, stackup and manufacturing review. Price and schedule are confirmed for your project.
