There is a fundamental tension at the heart of every RF receiver front end. The trade-off between selectivity and sensitivity sounds abstract, but it shows up in entirely practical ways the moment you start specifying a bandpass filter.
What the Terms Mean
Selectivity is the receiver's ability to discriminate between a wanted signal and an unwanted one at a nearby frequency. A highly selective receiver — one with a very narrow bandpass filter ahead of its LNA — rejects out-of-band interference before it can cause problems downstream.
Sensitivity is the receiver's ability to detect weak signals, limited primarily by noise figure and thermal noise floor. The filter's job is to keep unwanted power out of the LNA's input, preserving the sensitivity you designed into the system in the first place.
The Problem With Narrow Filters
Narrow filters are seductive. Tight bandwidth means excellent adjacent-channel rejection and clean isolation of the band of interest. In hardware, though, two problems emerge.
The first is insertion loss. Narrow filters have higher insertion loss than wider ones. Each dB of filter loss directly degrades the noise figure of everything downstream. A filter that costs you 3 dB has just raised the effective noise figure of your receive chain by 3 dB, regardless of how good the LNA is.
The second is stability. Every physical filter has a centre frequency that drifts with temperature. Dielectric constants, cavity dimensions, and coil inductances all shift as temperature changes. A filter that is perfectly centred at room temperature may be meaningfully offset at the temperature extremes your hardware actually sees in the field.
The drift may be small in absolute terms, but its impact scales with how narrow the filter is. A 500 kHz shift in a 20 MHz passband is a 5% displacement — tolerable. The same shift in a 2 MHz passband is 25% — you are now losing a significant portion of your passband to temperature alone. Ageing compounds this further as ceramic resonators and solder joints settle over years of operation.
The Cost of Going Wider
A wider filter tolerates the same drift with proportionally less impact, has lower insertion loss, and is less sensitive to manufacturing variation. The cost is that more out-of-band signal reaches the LNA. In a noisy RF environment, that can compress the amplifier, raise its noise figure, and generate intermodulation products that land directly on your wanted channel.
A Concrete Example: 315 MHz Receiver Design
The 315 MHz band used by automotive key fobs, garage door openers, and short-range remote controls is a good illustration of this trade-off in practice. The band sits alongside other ISM and unlicensed activity, and real deployments range from quiet rural environments to crowded parking structures where dozens of key fobs may be operating in close proximity.
We offer two bandpass filters centred at 315 MHz that bracket this trade-off directly. The 4 MHz bandwidth version gives you comfortable margin against thermal drift and manufacturing variation. A centre frequency shift of several hundred kilohertz barely touches the passband edges.
It suits applications where the hardware will see wide temperature swings, where the oscillator or filter components are not tightly temperature-compensated, or where the receiver needs to capture OOK or FSK signals that occupy a meaningful portion of that bandwidth. The trade-off is that interferers within a few megahertz of 315 MHz get through to the LNA, so front-end linearity matters.
The 500 kHz bandwidth version is a different proposition entirely. At that bandwidth, rejection of adjacent signals is dramatically tighter — useful in a dense interference environment where you need to isolate a narrow-channel signal cleanly.
But 500 kHz leaves almost no margin for drift. If your filter shifts even 200 kHz from its nominal centre frequency due to temperature, you have lost 40% of your passband. This filter belongs in designs where the hardware is thermally stable, where the signal chain is well-characterised over temperature, and where the interference environment genuinely warrants that level of selectivity. It is not a safe default for a product going into an unknown deployment environment.
The Right Answer Depends on Three Things
There is no universally correct bandwidth. The right answer depends on how stable your RF environment is, how stable your hardware is, and how much interference you are actually facing — not assuming worst case, but measuring it.
A filter specified without accounting for thermal drift works beautifully on the bench and degrades in the field. A filter specified without accounting for the real interference environment may be wider than necessary, trading selectivity for tolerance it does not need.
The best filter specification starts with a clear-eyed assessment of the environment the system will actually live in. If you are unsure which of our 315 MHz filters fits your application, we are happy to talk through the interference environment and thermal requirements before you commit to a design.