I Chose the Wrong Murata SAW Filter for a 2660 Flip Design (And What AIS 162 MHz Taught Me)

When I first started designing RF front-ends for ruggedized flip phones, I made a classic rookie error. I assumed that if a component—like a Murata SAW filter—was small and had the right center frequency, it was a safe drop-in. I was wrong. Really wrong.

The mistake happened on a project for a 2660 flip mechanism. A $3,200 order for a prototype batch. I specified a filter that looked perfect on paper. Three weeks later, we had a pile of dead boards. The culprit? A mismatch between the filter’s rejection bandwidth and the adjacent AIS (Automatic Identification System) band at 162 MHz.

This article is a walkthrough of that failure. I’ll compare two approaches: the 'quick-spec' method I used initially, and the 'system-aware' method I learned the hard way.

The First Dimension: The Core Assumption (Center Frequency vs. Real-World Interference)

The 'Quick-Spec' Mistake

I picked a standard Murata SAW filter. Its datasheet said it was for a specific receive band. The center frequency was correct. The insertion loss was excellent—under 2 dB. I was happy. Period.

The 'System-Aware' Reality

Here's the thing: the 2660 flip design put the antenna and the filter on the same flex cable. That cable acted like a secondary antenna for out-of-band signals. The biggest offender? The nearby AIS 162 MHz broadcasts used for ship-to-ship and ship-to-shore communication. Our device was near a port (this was back in 2022). My filter had okay rejection at that frequency, but not enough to handle the signal the flex cable was picking up.

From the outside, picking a filter with the right center frequency seems like a no-brainer. The reality is that in a tightly packed device, the 'right' center frequency is just the starting line. The real battle is against the unexpected interference sources your board layout itself creates.

The numbers said the filter would pass the intended signal. My gut (and two ruined prototypes) said the system was more complex. Trust the system, not just the datasheet.

The Second Dimension: PCB Layout vs. Filter Performance

The 'Quick-Spec' Approach

I placed the filter where it fit on the 2660 flip board. Simple. Done.

The 'System-Aware' Approach

This is where the real lesson lives. The Murata SAW filter’s performance is intimately tied to the impedance of the traces going in and out. On a 2660 flip design, space is at a premium. You’re forced to use long, thin traces. I didn’t model the parasitic capacitance these traces introduced. That capacitance shifted the filter’s rejection notch by nearly 15 MHz.

I’m not talking about a subtle degradation. I’m talking about the filter effectively not working as a blocker for the AIS 162 MHz signal. The rejection was out of band.

The initial approach was to blame the filter. The correct approach was to blame my layout. Lesson: the filter and the PCB are a single system.

The Third Dimension: Total Cost of Ownership (TCO) of a 'Cheap' Mistake

The 'Quick-Spec' Cost

The filter itself cost maybe $0.15 in volume. The re-spin of the 2660 flip board cost $890. The delay cost us a 1-week production cycle and an angry customer.

The 'System-Aware' Cost

Using a slightly larger filter (better rejection, lower proximity sensitivity) would have cost an extra $0.30 per unit. It would have also required a marginally different layout. The total premium: maybe $1.50 per board in the prototype run. A $1.50 solution to an $890 problem.

Look, I'm not saying the cheapest filter is always the wrong choice. I'm saying that when the mechanical constraints of your design (like the 2660 flip hinge and flex cable routing) force a specific layout, the 'cheap' component can become the most expensive mistake.

People assume the lowest BOM cost is the most efficient. What they don't see is the deferred cost of re-spins, testing, and field failures.

The Verdict: When to Use Which 'Type' of Murata Filter

Use a standard, high-performance filter (the 'quick-spec' type) when:

Your antenna and filter are on a rigid PCB with controlled impedance.
You have ample ground plane to isolate sensitive traces.
You know exactly what the interference environment looks like (i.e., you've done a site survey or your device is for a controlled environment).

Switch to a 'system-optimized' filter (think broader rejection or package-level shielding) when:

Your design uses flex cable (like the 2660 flip hinge).
The filter is placed near a known high-power interferer (e.g., AIS on 162 MHz).
You can't easily re-spin the board (short timeline, high tooling costs).

The elegant solution wasn't the one that fit on the first try. It was the one that acknowledged the compromise up front. The lesson? In RF design for compact devices, total cost of ownership includes the cost of your own future debugging time.