A holistic view of 77GHz radar sensors as a PCBA build, considering fabrication, assembly and packaging materials.
The Society of Automotive Engineers (SAE) and US Department of Transportation classify levels of vehicle autonomy from 0 to 5. Level 0 incorporates no automation; levels 1-3 have varying degrees of partial assistance to the driver, where the automobile, for example, can control steering, acceleration and deceleration, and even interfere with the driver. Finally, in full autonomy, level 5, the car drives on its own and makes all decisions and reactions to its surroundings.1
The automotive market uses a combination of sensors to make these critical decisions. Radar designs are the fastest growing sensors in ADAS today, due to the longer-range capabilities and their resistance to all weather conditions.2 This research will focus on radar designs, specifically long-range 77GHz radar, to showcase how automotive materials are changing and, through the choice of alternatives to those conventionally used in the space, how product life and reliability can be enhanced.
Basics of Construction
Circuit board materials. Starting with the backbone of the design, the system requires transmission and reception of radio waves through antennas to gather information on objects around the vehicle. This is made as exposed metal circuits on a dielectric substrate material. The length and width of the trace are specifically defined for the frequency that will be used to propagate the signal. Close attention is paid to trace height, width and shape. The critical nature of this part requires the information that goes out and comes in as a signal is not lost or altered. To ensure minimal disruption, the antenna is created on a “low loss” substrate material. The term low loss refers to dissipation factor (Df), described in physics as a loss rate of energy of oscillation.3 This is exactly the technology used in airplanes to inform pilots of their surroundings, especially when visibility is limited.
The European Telecommunications Standards Institute (ETSI) and Federal Communications Commission (FCC) developed regulations to phase out 24GHz use for automotive applications. These regulatory changes and the need for increased performance have caused a shift to 77GHz.4 The increase in frequency requires a change to the substrate used. As frequency increases, more signal is carried on the outer edges of the conductor. This is known as skin effect. As a signal approaches the edges of the conductor, there is more opportunity for dielectric loss due to competing interactions within the design. This includes circuits in proximity to each other.
The 77GHz designs require enhanced low-loss dielectric materials to achieve the level of reduced loss over their 24GHz counterparts. In response, substrate suppliers created dielectric made of polytetrafluoroethylene (PTFE) that are ceramic-filled. These are designed to have specific electrical stability, very different from the traditional resin-filled glass-weave substrates (FR-4) the automotive industry uses for other applications, including infotainment, body control or even engine control. Being in the radio-frequency range requires a very low dissipation factor. By comparison, the Df value of FR-4 is 0.20, while that of ceramic-filled PTFE is 0.002. Yet, there are challenges. These materials cost more and require altered fabrication processes. Due to the critical nature of the safety application, ceramic-filled PTFE is by far the most heavily used for 77GHz radar safety designs.
Polytetrafluoroethylene substrates are well established materials for military, aerospace and some telecommunications applications, but care should be taken when processing, especially when considering the volume these designs will approach as autonomy levels increase. The material is deformable, which means it can scratch or bend easily with handling. It is highly chemically resistant, requiring different cleaning and conditioning preparation steps prior to copper plating. Additionally, ceramic-filled PTFE has a very high porosity that can and will absorb solution throughout the PCB fabrication. Some chemicals will attach to fillers in the substrate. Special attention is to be paid to the process cycle parameters to successfully manufacture the precision required for antenna designs. It is recommended to follow the suggestions detailed by the substrate supplier.
The base material is the first source of loss associated with the design. A second opportunity for loss is on the conductor or the metal circuit formed as the antenna. The conductor in this work is copper. The fabricator receives the substrate as a thin layer of copper foil, which is printed and etched, then plated to the desired thickness with electrolytic copper plating. The width, height and roughness of that copper is important for signal propagation. The final step of the fabrication process is to coat the copper with a finish that will resist oxidation and preserve the rest of the design prior to assembly, protecting the exposed antenna in end use. Conductor loss can occur depending on the type of surface finish chosen. Here, the types of surface finishes and their effect on insertion loss will be discussed. The testing will determine if coating thickness plays a role in performance consistency.
Assembly materials. The function of an electronic system is driven by semiconductor packages attached to the surface of the PCB. (Details of the packages are discussed in the next section.) These components are joined to the circuit board using a solder alloy. Most assemblies are built by printing a solder paste, placing components on that paste, and conveying through a heated oven to create an intermetallic bond. During this process, flux material prepares the surface for attachment and is volatilized, leaving behind the metal alloy and certain flux residue. The resultant solder joint creates a mechanical bond and an electrical pathway between the component and PCB.
Historically, the automotive market primarily used SnPb solder. In 2010, regulatory changes pushed adoption of lead-free (tin-silver-copper or SAC) alloys.
The most obvious challenge for a solder alloy (in automotive applications) includes harsh operating conditions, as devices operate close to the engine and are exposed to extreme weather conditions. Concerns surrounding product life and reliability are directly connected to the environmental exposures and the expected high-performance processing and function required for life-critical functions.
It is well-documented that thermomechanical forces put stress on the solder joint. Heating the solder joint is realized in two ways, from the function of the design and/or the location of the systems. For example, devices close to the car engine can see operating temperatures approaching 150°C. Thermomechanical stress can result from multiple aspects, such as vibration of the unit or exposure to temperature fluctuations. The most common in all electronic systems is movement of dissimilar materials due to heat. The dissimilar rates of expansion between a stiff component and that of a PCB create a coefficient of thermal expansion (CTE) mismatch that degrades the solder joint over repetitive temperature cycles. Ultimately, the majority of stress from everything moving is directed to the solder joint, which then fractures, rendering a loss of electrical contact from component package to the circuit board.5
When an alloy is exposed to prolonged thermal or mechanical stresses, it will experience a deformation called creep. Creep will occur at elevated temperatures when the atoms are most mobile.6 In the early 2000s, an industry consortium was created to address the challenges in automotive electronics as a result of greater environmental stresses and increased component performance needs. Car makers, tier ones, and material suppliers organized a consortium to develop new alloys that could improve thermal resistance, specifically thermal cycling, while maintaining SAC 305 reflow temperatures. The resulting alloy, Innolot, was patented in 2003 and comprised a SAC-based material with alloying additions to alter the microstructure and enhance creep properties at higher operating temperatures (TABLE 1).5
Alloy development incorporates four major areas of focus: solid solution strengthening, diffusion modifiers, grain refinement and precipitate strengthening. Differences in composition will result in microstructure difference, which in turn affect mechanical properties. This work will further illustrate that the Innolot solder alloy will extend the life of a product, even when exposed to significant thermal stresses. Testing will include the surface finishes examined in the insertion loss study mentioned above.
Packaging materials. As mentioned, the function of any electrical system comes from the IC substrate packages. The final area of discussion for this work focuses on material changes to enhance the reliability of these components. In radar sensor designs, the components generate the radio frequency, direct the energy, and identify the information that is retuned. Moving from 24GHz to 77GHz brings greater resolution to the information gathered surrounding the car. The increased power and processing require a change to the chosen package. Designs supporting 24GHz use quad flatpacks that contained a low input/output (I/O) count, on the order of 32 pins.
The 77GHz designs require increased I/O and, with that, a transition to a ball grid array (BGA) package with about 180 connections. BGAs offer several advantages over leaded components, such as increased density, but without the concern for solder bridging during assembly and greater heat conduction. Heat can travel more easily to the PCB on a BGA. This decreases the heat experienced by the silicon chip during its operating life. Last, and likely most important for the 77GHz radar sensor, the BGA provides low inductance. The short distance between the package and PCB reduces signal loss or distortion.
This main BGA used to function the antenna is called an MMIC or monolithic microwave integrated circuit. In the construction, the die is attached to the substrate using a solder alloy or a conductive adhesive. The final portion of this study will investigate a thermal conductive adhesive attachment of die to the substrate as a tool for thermal management of the die. Pulling heat from the die will extended the package life. Thermal conductivity will be measured to understand the material differences. In addition to longer package life, adhesive is another avenue to eliminate lead in automotive designs. Highly filled silver adhesives, such as the hybrid silver sintering die used to attach materials, can offer performance benefits above and beyond solder materials.
One of the first areas of concern is getting a previously stored package through standard assembly processing. The defect of concern is delamination of the epoxy mold compound from reflow exposure. During storage, packages can absorb moisture prior to assembly. There is a risk the heat from reflow will volatilize the moisture, thereby putting pressure on the epoxy and pushing that epoxy from the rest of the package, resulting in a defect known as “popcorning.”
All packages are rated on a moisture sensitivity level grading as documented by J-STD-020D, a moisture sensitivity classification for non-hermetically-sealed surface mount devices (TABLE 2).7 End-users want to ensure components can be stored for an extended time prior to assembly and function reliably in the final build. Although more of a concern for leadframe packages, attention should be paid to all larger devices, especially for critical applications. This testing will ensure changes made for improved thermal transfer from the die will not adversely affect overall package construction.
Automotive Test Conditions
The automotive industry has always relied on accelerated tests such as thermal cycling in air to predict the quality of a design and its model’s ultimate product life expectancy. Principles derived from the Coffin-Manson equation enable predictive modeling of solder joint lifespan by accounting for the effect of plastic deformation of a material over a specified temperature range and dwell time.6 The resulting equation predicts the rate of deformation of the solder joint will increase as temperature extremes become farther apart and cycle time increases. Radar designs serve a critical role in safety as the automotive industry moves toward complete vehicle autonomy. As such, the performance requirements of the devices are held to a higher standard. Market-defined levels for high- and low-temperature ranges are typically based on expected operating temperatures of the device. For example, interior systems are tested to a lower peak temperature than those for engine control, simply due to the operating environment. Safety systems such as radar start to blur the lines because they may or may not be located in harsh areas. Also, the duration or length of the accelerated tests have been increased for safety systems due to their critical nature.
To simulate the environments of automotive electronics, accelerated tests are used specific to functioning electronics. They include thermal cycling and sensitivity to moisture.
Test procedures. For this experimentation, each material change was tested separately to understand its full performance improvement or lack thereof. For a production build, it is recommended to test the parts both separately and as a full construction to understand improvements that can be realized for a specific application. To test the various levels of the PCBA, automotive industry methods were used. All reflow processes were held to the J-STD reflow profile, which is considered a more aggressive profile due to the long soak zone and high peak temperature. For all testing, a peak temperature of 250°C was used. Thermal cycling was tested against an aggressive thermal cycling profile of -40° to 160°C, with 10 min. dwell for 2,000 cycles (TABLE 3).
Surface finish effect on insertion loss. To understand insertion loss differences between surface finishes, the microstrip differential phase length method was used. Microstrip test vehicles created from a leading PTFE substrate material specifically used for 77GHz radar designs were fabricated in a production environment. The strips were processed in pairs measuring 50.8mm and 203.2mm (2" and 8") length, each having a width of 0.75". For ease of handling, the 5-mil thick PTFE-based material was adhered to an FR-4 backing layer for rigidity. The test vehicles were plated with the following production-available surface finishes: high-temperature organic solderability preservative (OSP), immersion silver (ImAg) and immersion tin (ImSn). Each finish was coated to both a thin and thick level following technical datasheet process parameters and conditions (TABLE 4). This was chosen to illustrate any effect of oxidation due to coating porosity for the OSP and immersion silver, or any influence of the immersion tin intermetallic compound growth. In addition to the microstrips, specific thickness coupons were added for separate thickness measurements based on the requirements for each surface finish. These included strip and UV absorption of the OSP, and coulometric reduction of the tin, to determine pure tin versus intermetallic compounds that may have formed under conditioning.
After coating the surface finishes, thickness was determined on thickness test coupons using a UV dissolution method for the OSP, x-ray fluorescence for the silver and coulometric reduction for tin.
Microstrip parts and additional thickness coupons were sent to the Rogers Corp. research facility for insertion loss measurements before and after two exposures to a reflow profile with peak temperature at 250°C in a 7-zone convection reflow oven.
Insertion loss measurements were taken for each circuit across a wide range of frequencies on a Keysight model #N5251A millimeter-wave network analyzer. Insertion loss data were gathered as-coated and after two exposures to the reflow oven. Samples were then returned for post-reflow thickness measurements and appearance observation.
Solder alloy effect on solder joint integrity. The initial investigation of an alternate alloy to SAC in the early 2000s was to improve solder joint reliability for high-temperature applications. Although radar systems are not considered to have the highest operating temperature in a vehicle, they are tested to extreme peak operating conditions to effectively stress the solder alloy and distinguish its performance characteristics.
An active and passively monitored test vehicle was used for the solder joint integrity evaluation. The active test vehicle included 15 BGA-84 components daisy-chained for in-situ resistance measurements throughout the thermal cycle conditions. The second test vehicle consisted of passive components such as 1206 and 0805 resistors for shear strength testing measurements at defined intervals. Shear strength testing is a commonly accepted indicator of solder joint reliability, as it relies on the assumption mechanical reduction of a solder joint over thermomechanical aging leads to failure of the electronic circuit. For this work, the same surface finishes used for insertion loss were tested in conjunction with the solder alloys. Only the upper thickness levels were used to reduce the test size. It should be noted this work only includes a portion of the fully executed solderability testing. More information can be shared upon request.
In total, 15 components per test vehicle and three test vehicles (FIGURE 2A) per condition were measured. Including three surface finishes in the evaluation resulted in a total of 540 measurements taken. The BGA designs were arranged in a daisy chain to allow in-situ measurements. Failures were identified as a condition where electrical resistance of the device increased in excess of 20% of initial resistance for five consecutive readings per IPC-9701A.
Shear strength was analyzed before and after 1,000 and 2,000 thermal cycles. Shear testing was executed on the 1206 resistor for the “automotive test vehicle” in FIGURE 2b. Parts were tested for high-speed shear (1,000µm/sec.) evaluation at 0, 1,000 and 2,000 cycles to estimate the effect of thermal cycling on solder joint strength. The test vehicle is an 8-layer board and includes multiple BGAs, LGA, MLF and resistors for evaluation.
Material choice for enhanced thermal management from the die. As device complexity increases and motor vehicles are asked to do more, the subsequent increased processing power leads to increased heat generated.8 Removal of this heat from the die within the IC substrate will promote improved performance of the package and extended life. Standard ICs use materials that remove 2-10W/m-K. Today, greater thermal transfer is a necessity. For advanced safety systems such as the MMICs used in the radar designs, the desire is greater than 10W/m-K. It is common to use a thermally conductive adhesive for heat transfer from die to leadframe. A new thermal conductive adhesive available can be used with metal or substrate attachment. The material combines micron silver flake, thermoset resins and diluents to form the die connection.
To prove the performance of this material, thermal conductivity was measured at 175°C for a 60-min. exposure. Laser flash equipment was used to measure the thermal conductivity of the die attach material. The effective thermal conductivity (K-eff) is typically measured in a multilayer format. This enables the analyst to understand all factors of contact resistances at the interfacing surfaces.9 A plastic BGA measuring 7mm x 7mm was used to test the performance of the thermally conductive epoxy in comparison to the conventional material. In a production environment, the material was ink-jetted in an “x” pattern on the substrate, followed by die placement. The epoxy was cured at 175°C for 60 min. The die was then wire-bonded, and the package was completed with final encapsulation. Parts were analyzed by x-ray for any evidence of delamination after MSL 3 preconditioning.
Surface finish effect on insertion loss. In agreement with previous work executed by Rogers, the OSP and immersion silver resulted in the same insertion loss as the bare copper control10 when analyzing from 0 to 80GHz. The graphs for both finishes overlay regardless of coating thickness or reflow conditioning. Immersion tin had the greatest loss compared to the bare copper control, and as the frequency increased, the delta between the two became greater (FIGURE 3). Tin also showed a significant difference when comparing as plated to post-reflow conditioning.
When focusing on the 77GHz measurement, the insertion loss comparison between finishes becomes clearer (TABLE 5). The tin insertion loss is about 0.2dB/in greater than the bare copper, OSP and immersion silver. This might not be a concern for lower frequencies, but it becomes a concern for systems operating at higher frequencies. In addition, the thicker tin deposit has greater losses after reflow conditioning.
It is known immersion tin thickness changes with heat exposure, which is not a characteristic for OSP or immersion silver. Pure tin was measured on the same coulmetric equipment to determine thickness change after the two reflow conditions. Three readings per sample were taken to calculate an average. The sample initially coated with 1.0µm of pure tin averaged 0.2µm remaining, which results in an estimated intermetallic compound (IMC) layer of 0.8µm. The samples plated to 1.3µm had 0.35 remaining pure tin and an estimated IMC layer of 0.95µm. The changes in insertion loss could be attributed to the greater amount of IMC compound. Additional work will be done to get a more precise answer for why the tin performance is different with reflow conditioning.
Only the thin OSP coating displayed a change in appearance after reflow exposure. It can be described as a slight darkening with more iridescence. Even with the slight change in appearance, there was no detriment to insertion loss. Immersion silver did not experience any appearance change, and no change in insertion loss was observed. The immersion tin also did not display any visible appearance change, but the change in insertion performance was observed over the range of frequencies tested. Proper choice of surface finish is important at 77GHz frequencies.
Solder alloy effect on solder joint integrity. The strongest factor found to influence thermal cycle resistance was the solder alloy. For all surface finishes tested, the Innolot alloy exhibited the highest resistance to cracking and thus lowest propensity to change in electrical resistance over continual thermal cycles, outperforming the SAC 305 baseline by over 25% on average. The -40º/160ºC range was specifically chosen to effectively evaluate the creep-resistant properties of the solder alloys. Given predictive modeling, lower temperature extremes would result in higher cycles to failure. Additionally, given the failure criteria as defined by IPC-9701A, a change in resistance of greater than 20% does not indicate the resultant device is no longer electrically functioning.
The interaction (FIGURE 4) and resulting Weibull plots (FIGURE 5) suggest minimal variation among the finishes, with immersion tin exhibiting the highest propensity to change in resistance. Results suggest the amount of free tin affects the failure rate due to continual exposure to higher temperature ranges. Further investigation of these results will be examined in future work. The immersion silver and Cu OSP surface finish resulted in 63.2% of the parts tested passing resistance readings above 1,750 cycles, as illustrated by the Weibull plots for the Innolot alloy on the challenging thermal cycle profile described above.
The passive component test vehicle results aligned with the results seen on the actively monitored BGA devices. As expected from the sum of squares calculation, shear strength of both alloys was affected most by the number of temperature cycles (FIGURE 6). However, the Innolot solder alloy exhibited the lowest rate of shear strength reduction over 2,000 thermal cycles relative to the SAC 305 baseline. The results indicate the design of the Innolot alloy is more resistant to mechanical reduction over the number of temperature cycles. Similar to the active device test vehicle, shear strength was found to be consistent across each of the surface finishes evaluated (FIGURE 7). In each of the instances, a tin-rich Cu6Sn5 intermetallic is formed. Ultimately, regardless of alloy or surface finish, the largest rate of reduction in shear strength occurs between 0 and 1,000 thermal cycles, with the Innolot alloy maintaining a significantly higher shear strength over 2,000 cycles relative to SAC 305.
Thermal management of die. Bulk conductivity does not accurately predict the thermal resistance of a package. It does not consider interfacial resistance between the die and adhesive. The previously developed “effective” conductivity test resulted in a dramatic improvement from the high silver and hybrid silver adhesives. Compared to a conventional epoxy die attach, the hybrid silver sinter material showed was almost three times greater, at 14W/m-K compared to 5W/m-K (FIGURE 9). As this was demonstrated on a bare silicon die, another advantage is realized compared to leaded solder materials and eutectic gold-tin. It does not require back-side metallization.
A PBGA measuring 7mm x 7mm was used to test MSL resistance at 260°C. Parts were checked by x-ray for delamination after preconditioning to MSL 3 exposure. No evidence of delamination or voiding was observed (FIGURE 10). The new adhesive does not pose a threat to the required epoxy mold adhesion.
Significant advances are happening in the automotive space to deliver greater safety to passengers and pedestrians around the vehicle. These enhanced safety requirements utilize specialized technologies with enhanced requirements to achieve success. As this research demonstrates, it is imperative the proper surface finish be chosen for high-frequency applications. Enhanced solder alloys such as Innolot are a requirement to withstand the aggressive thermal cycle conditions required for advanced safety. Innolot in this testing shows a dramatic increase in thermal cycle resistance with superior shear strength. Finally, the hybrid silver sintering adhesive will also extend product life by improving the thermal transfer away from the die.
Each of these materials provides improvements over the incumbent but could realize even greater improvement when used together in one design. It is critical to test individual material changes, as well as those combined in the final design build for a complete understanding of end-use life and performance.
This work was a group effort. I’d like to extend appreciation to all those involved, including the MacDermid Alpha circuitry, assembly and semiconductors groups, the Rogers’ team and my marketing team for all their help.
This article is adapted from a paper published in the IPC Apex Expo Proceedings in February 2020 and is published here with the authors’ permission.
1. Synopsys, Dude, Where’s My Autonomous Car? The 6 Levels of Vehicle Autonomy, company paper, synopsys.com/automotive/autonomous-driving-levels.html.
2. Avinash Kondepaddy, Why Are Automotive Radar Systems Shifting to 77GHz, Mar. 25, 2019, company paper, pathpartnertech.com/why-are-automotive-radar-systems-shifting-to-77ghz.
3. Wikipedia, retrieved May 11, 2020, wikipedia.org/wiki/dissipation_factor.
4. John Coonrod, “Understanding When to Use FR-4 or High-Frequency Laminates,” IPC Apex Expo, April 2011.
5. Harry Bhdeshia and Mike Ashby, “The Creep of Solder,” 2004, phase-trans.msm.cam.ac.uk/2004/creep.practical.pdf.
6. Craig Hillman, Ph.D., Reality of New, High Reliability Solders, DfR monthly webinars, Feb 24, 2016.
7. IPC J-STD-020D, “Moisture/Reflow Sensitivity Classification for Non-hermetic Solid State Surface Mount Devices,” August 2007.
8. Cypress Semiconductor application note, Understanding Temperature Specifications: An Introduction, cypress.com/file/38656/download.
9. H. H. Jin, S. Kanagavel and W.F. Chin, “Novel Conductive Paste Using Hybrid Silver Sintering Technology for High Reliability Power Semiconductor Packaging,” ECTC, June 2014.
10. John Coonrod, “The Impact of Final Plated Finishes on Insertion Loss for High Frequency PCBs,” SMTA International, September 2017.