Before this:Gain, AGC & avoiding overloadHow an SDR receiver works
Front-end filters, LNAs & overload
Key takeaways Gain is not the only thing that decides a clean decode — what you let into the front end matters just as much. A preselector/filter rejects out-of-band energy before it can do harm; an LNA lifts weak signals with minimal added noise; but a strong nearby transmitter can still cause overload, intermodulation (false ghost signals), or reciprocal mixing (a clean-looking carrier that decodes badly). The fixes are often less gain, an attenuator, or a filter — not more amplification. Builds on gain & AGC and the SDR receiver.
The gain lesson got your signals positioned correctly in the ADC’s range. But gain treats everything in the band the same way — it can’t distinguish your target from the megawatt FM station down the road. This lesson is about the analog stages before digitising, and the subtle ways a strong signal you aren’t even trying to hear can wreck the one you are.
The front end, in order
The SDR receiver lesson walked the whole chain; here we zoom in on the analog stages that come before the samples exist:
antenna → (filter) → LNA → mixer → ADC
The antenna collects everything in its range. An optional filter rejects frequencies you don’t want. The LNA (low-noise amplifier) boosts what’s left. The mixer shifts your band down to a rate the converter can handle, and the ADC digitises it into IQ samples. Every stage before the ADC is analog and shared — whatever energy reaches a stage affects everything passing through it, not just your channel.
Preselector & bandpass filters
A preselector is a filter placed early in the chain that passes the band you care about and rejects the rest. Why bother, when you’re going to filter digitally later anyway? Because the analog front end has to survive the total power hitting it, and much of that power is out of band: FM broadcast (88–108 MHz), pagers, cellular, and TV transmitters are often tens of dB stronger than the trunked control channel you want.
- A bandpass filter passes a range and attenuates everything outside it.
- A notch filter does the opposite — it kills one troublesome band (an FM-broadcast reject notch is the classic first purchase).
- SAW filters (surface acoustic wave) are compact fixed-frequency bandpass parts common in ready-made SDR filter modules.
A filter can’t add signal — it only takes energy away, and it costs a little of your wanted signal too (insertion loss). What it buys you is protection: everything downstream now sees a much smaller total power, so the LNA, mixer and ADC are far less likely to overload.
Low-noise amplifiers (LNAs)
An LNA amplifies weak signals while adding as little noise of its own as possible. It matters where it sits because the first amplifier dominates the system noise figure — noise added at the front is amplified by everything after it, so a quiet first stage sets the floor for the whole receiver. This is why a masthead LNA (mounted at the antenna) can help so much: it lifts the signal before the loss of a long feedline drags the SNR down.
That leads straight to the classic ordering trade-off:
- Filter before LNA — the filter protects the amplifier from strong out-of-band signals that would otherwise overload it. The price is the filter’s insertion loss, which adds directly to the noise figure.
- LNA before filter — the best possible noise figure, because nothing lossy sits ahead of the amplifier. The price is that the LNA is exposed to the full band and can itself be driven into overload.
In a quiet rural location, LNA-first wins. In a crowded urban RF environment with strong nearby transmitters, filter-first is usually the safer choice — a slightly higher noise figure beats an amplifier that’s constantly overloading.
Overload: too much total power
The mixer and ADC don’t see your channel in isolation — they see the sum of everything in the band at once. A strong nearby transmitter can drive them into compression (the stage stops responding linearly) even when your target is weak and far away in frequency. Once a stage is compressed, it distorts every signal passing through it, including yours.
This looks similar to the too-much-gain problem from the gain lesson, but the cause is different: with clipping you turned the gain up too far, and turning it down fixes it. With front-end overload the input is too hot — the culprit is an external signal, and the fix is to reject that signal or attenuate the whole input, not just to trim gain. Watch for it whenever a channel decodes fine at some times and falls apart when a strong local transmitter keys up.
Intermodulation (IMD)
When a non-linear stage is fed two strong signals at f1 and f2, it doesn’t just distort them — it mixes them and produces new intermodulation products at frequencies like 2f1−f2 and 2f2−f1. These third-order products fall close to the original signals, so they can land right on top of a real channel you’re trying to decode, and they look exactly like genuine signals on the waterfall.
The giveaway is how they behave with level. A real signal barely changes when you add a few dB of attenuation. An intermodulation product changes much faster — a third-order product drops about 3 dB for every 1 dB you attenuate the input. So if a suspected ghost signal vanishes when you insert 10 dB of attenuation while your real signals only dip slightly, you’ve found IMD, not a station.
Reciprocal mixing & phase noise
This is the subtle one. A real local oscillator isn’t a perfect single tone — it has phase noise, a skirt of noise spread either side of its frequency. In the mixer, your wanted channel isn’t the only thing that beats against the oscillator: a strong nearby signal does too. When that strong signal mixes with the oscillator’s phase noise, it drags a copy of that broadband noise skirt right onto your weak channel. This is reciprocal mixing.
The trap is what it looks like. On a wideband FFT the strong carrier looks perfectly clean, and your channel looks fine — there’s no obvious spur, no clipping. But the demodulated signal is degraded: the effective noise on the channel is raised and the EVM climbs, so it decodes poorly despite looking healthy on the spectrum. The strength of the offender and the quality of the oscillator set how bad it gets.
More gain does not help here — it amplifies the smeared noise right along with the signal. The fixes are a cleaner front end (a better, lower-phase-noise oscillator) or knocking the strong offender down with a filter so it never reaches the mixer at full strength.
Fixes: attenuators & filters
The counter-intuitive part of front-end trouble is that adding an attenuator — or lowering gain, or fitting a bandpass or FM-reject filter — can improve your decode. You’re trading a little signal for a large reduction in the total power abusing the front end, and clearing overload or IMD wins that trade easily. A practical checklist when a channel decodes worse than its raw strength suggests it should:
- Check the gain first — rule out plain ADC clipping.
- Insert 10 dB of attenuation. If SNR improves or ghost signals vanish, you were overloaded — the front end, not the channel, was the problem.
- Identify the strongest offender on a wide spectrum (often FM broadcast or a pager); fit a notch or bandpass filter to reject it.
- In a crowded location, move the filter ahead of the LNA.
- Re-check level and SNR on the tuning meters — a good change shows as a higher, steadier SNR, not just a lower level.
Quick check: ghost signals appear across the band and vanish when you add 10 dB of attenuation, while your real signals only dip slightly. What is it?
Recap
- The analog front end — antenna → filter → LNA → mixer → ADC — decides what even reaches your decoder, and every pre-ADC stage is shared by the whole band.
- A preselector/filter can’t add signal, but it rejects out-of-band energy (FM, pagers, cellular) and protects everything downstream from overload.
- The first amplifier dominates the noise figure: LNA-first for best sensitivity, filter-first to survive a crowded RF environment.
- Overload (too much total input power) and intermodulation (false products like 2f1−f2) are caused by strong signals, not by your gain — an attenuator often cures them.
- Reciprocal mixing raises the effective noise on a weak channel via oscillator phase noise; the carrier looks clean on the FFT yet decodes badly, and more gain won’t help — a cleaner front end or a filter will.
Next up: the dongles all of this runs on — SDR hardware: RTL-SDR, HackRF & Airspy.
Frequently asked questions
Why do I need a filter if I already have a gain control?
Gain decides how strong everything is when it reaches the ADC, but it can’t tell your target apart from a strong out-of-band transmitter — it amplifies both equally. A filter rejects the unwanted energy before it can overload the front end, so the parts of the chain that follow only see the band you care about. Gain and filtering solve different problems; a busy RF environment usually needs both.
What is intermodulation?
When two or more strong signals hit a stage that isn’t perfectly linear (a mixer or an overdriven amplifier), the stage mixes them and produces false products at new frequencies — combinations like 2f1−f2 and 2f2−f1. These land on real channels and look like genuine signals, but they’re artifacts of overload. The tell is that they appear and disappear as you change the input level with an attenuator.
Do I put the filter or the LNA first?
It’s a trade-off. A filter before the LNA protects the amplifier from strong out-of-band signals that would overload it, at the cost of the filter’s insertion loss adding to the noise figure. An LNA before the filter gives the best noise figure — the first amplifier dominates system noise — but leaves that amplifier exposed to everything. In a clean location, LNA-first; in a crowded RF environment, filter-first.
Why does a strong FM or pager signal ruin a channel far away from it?
A very strong nearby transmitter can push the mixer or ADC into overload even though it’s nowhere near your frequency, spraying intermodulation products and a raised noise floor across the band. It can also degrade your channel through reciprocal mixing — the strong signal mixes with the local oscillator’s phase noise and smears broadband noise onto your weak channel. The cure is to reject the offender with a filter, not to add gain.