In RF design, should you choose a BAW or a SAW filter?

Two patch filters with similar appearance are in front of us: the package size can just fit into the board, and the center frequency looks quite appropriate. At this time, if the decision is made only because they are printed with “BAW” or “SAW” tags, the key factors that really affect the performance of the RF link are often ignored: whether the passband edge is steep enough? Where does the interference in the adjacent frequency band fall? How is the transmit power calibrated? And the most important point-after this thing is actually soldered to the PCB, can it still maintain the good performance written in the specification?

To really compare BAW and SAW, they must be viewed at the same system boundary, rather than simply taking the technical name as the final conclusion.

Let’s first look at how these two structures achieve frequency selectivity

The SAW (surface acoustic wave) filter utilizes an acoustic wave propagating on the surface of a piezoelectric substrate. The input of the interdigital transducer (IDT) first converts the radio frequency voltage into a surface acoustic wave, after a period of propagation and interference, the output of the transducer and then the acoustic wave reduction to electrical signals. The period of the electrode, the sound wave propagation velocity of the material, the aperture size, the number of fingers and the reflection structure, these parameters together determine its operating frequency, bandwidth, insertion loss and suppression characteristics.

In contrast, the acoustic resonance of a BAW (bulk acoustic wave) filter occurs mainly in the thickness direction of the piezoelectric film, or inside the material. Common practices are film bulk acoustic resonators (FBAR) and firmly mounted resonators (SMR). Although it belongs to the same acoustic filtering technology as SAW, the energy propagation direction and resonant structure of BAW are different. This structural difference does affect the frequency band, Q value, power capacity and heat dissipation performance that the device can achieve, but this does not mean that as long as it is BAW, its passband or the reliability of a specific model must be better than SAW.

Don’t blindly select devices just by looking at the frequency band

SAW is very suitable for small size, large quantities of fixed frequency filtering, in mobile phone communication, GNSS navigation, remote telemetry, broadcasting and Internet of Things RF front-end in particular. BAW, on the other hand, is usually more advantageous in those occasions that require high frequency, high Q value, better temperature stability or can carry high power. However, with the advent of temperature-compensated SAW(TC-SAW) and other advanced SAW technologies, the boundaries of their traditional applications have long been less distinct. Therefore, the outdated notion that “high frequency necessitates BAW” is actually overly simplistic.

In the selection, be sure to pick the actual passband details, including the upper and lower edge frequency, center frequency, bandwidth and the corresponding reference point. Then, sort out which bands must be suppressed. It should be noted that the stopband suppression must be implemented to a specific frequency or frequency band; even if there is an extremely deep attenuation valley (notch) in the specification, it cannot represent the minimum suppression level of the entire stopband. If the roll-off at the edge of the passband is particularly steep, it is safe to take into account the frequency drift caused by the change of operating temperature and the tolerance of the device itself.

Don’t forget to put the loss and signal integrity back into the link to consider

The front end of the receive link is extremely sensitive to insertion loss. When making a comparison, it must be ensured that the two are measured under the same temperature, impedance, matching network and reference surface conditions. Focus on the “maximum loss” in the entire passband, don’t just stare at the good-looking “typical values” picked out by various manufacturers “. In addition, return loss, standing wave ratio (VSWR) and port impedance must be combined, otherwise you can easily figure out whether this is the problem of the device itself or whether the board level matching is not done well.

If you are running broadband modulation or a signal that is very sensitive to phase, you must also check the in-band fluctuation, group delay and flatness of group delay. A passable frequency response does not necessarily guarantee that the waveform fidelity will meet the requirements. If there are several stopbands listed in the data table, be sure to check them one by one with the actual interference source and the position of adjacent channels in the system.

Talking about power without considering the test conditions is meaningless

The maximum input power noted by the acoustic filter may refer to continuous wave power, peak power, full life average power, or even measured with a specific modulation waveform. The dBm or W values obtained under different test standards cannot be directly compared. In the transmission link, it is also necessary to carefully evaluate the temperature rise, frequency drift, harmonics, intermodulation, mismatch and service life. The device has not been burned in a short time, which does not mean that the system can work continuously for a long time at this power.

At high power, the SAW interdigital electrode may heat up, and its power bearing capacity will also be dragged down by the material, structure and packaging. The same is true for BAW. You cannot blindly believe that nominal power figure; thin films, acoustic reflective layers, or cavity structures are all affected by packaging stress and manufacturing deviations. The most reliable approach is to list the target waveform, average and peak power, temperature range, and mismatch conditions in the worst case in black and white in the acceptance criteria.

Final verification of port, package and board level implementation

First of all, confirm whether the device is a single-ended or differential port, what is the nominal impedance, whether to add an external matching network, and how to arrange the direction of package grounding and pads. When evaluating the performance of the differential system, it is absolutely impossible to directly copy the 50 ohm single-ended test results with convenience. In addition, parasitic parameters brought by pads, traces, grounding and packaging will also make the actual frequency response of the whole board deviate from the characteristic curve of the device itself.

When reading the SAW specification, check the balanced/unbalanced port, absolute attenuation, temperature range, input power, and anti-static (ESD) level. When reviewing BAW devices instead, in addition to the conventional S parameters, it is necessary to further ask about its power tolerance, temperature drift, harmonics, intermodulation, ESD performance and package stress conditions. When comparing, both parties have to take the complete specific material number and the corresponding version data to speak, and never use a broad technical pronoun to fool the verification of the material number level.

Pull all the alternative devices into the same specification table: after setting the passband, stopband, power, temperature and port conditions in a unified way, compare their maximum insertion loss, minimum suppression, return loss, group delay, size and packaging form. If the performance index of one of the options is “brushed” out with different test conditions, first fill in the gap data and make it aligned and comparable, and then decide whether this list is for SAW or BAW.