Part 9 of The Analog Edge, a 14-part field guide to the analog half of a GopherTrunk system. Part 8 audited our marginal reader’s feedline and found ~8 dB dying in forty feet of RG-58 and an adapter stack. New cable clawed back six of them. This part is about the next lever — active gain — and its sharp edge: an LNA in the wrong place amplifies nothing useful, and an LNA on a hot band manufactures the exact intermod garbage Part 4 taught you to recognize. Gain is easy; *clean gain is placement and filtering.*
TL;DR: An LNA helps exactly in proportion to how much loss comes after it, because the Friis cascade makes the first stage’s noise figure dominate the system. Mast-mounted LNA ahead of the coax: the run’s loss nearly vanishes from the noise budget. Desk-mounted LNA after the coax: you amplify signal and accumulated noise together and gain almost nothing. The tax is dynamic range — every dB of LNA gain is a dB less headroom against the FM broadcast / pager blowtorches, so on a hot band the LNA needs a filter riding with it. Power rides the coax via a bias tee (
bias_tee: trueon the device block for SDRs that provide one).
Key takeaways
- Placement beats specification. A mediocre LNA at the antenna outperforms an excellent LNA at the desk, because loss after amplification barely counts while loss before it counts dB-for-dB.
- Gain without filtering is a bet that your band is quiet. An LNA lifts every FM broadcast, pager, and cellular carrier in its passband along with your target; if the sum approaches the front end’s limits, you’ve traded noise floor for intermod.
- Filters spend dB to buy headroom. A bandpass or FM-notch filter has insertion loss — position decides whether that loss costs noise figure (before the LNA) or almost nothing (after it), and overload decides which you need.
- The decode-quality sweep is still the referee. Whether an LNA/filter combination netted out positive is answered the same way as gain staging in Part 3: by decode error rate, never by how loud the waterfall got.
Cheat sheet
| Concern | What it does | Where it lives |
|---|---|---|
| LNA fundamentals | sub-1 dB NF first stage; sets the cascade | low-noise amplifier, best SDR LNAs |
| Friis cascade | why stage 1 dominates system NF | noise figure |
| FM broadcast notch | kills the 88–108 MHz blowtorch before it intermods | FM broadcast filter, SDR filters guide |
| Band-specific filtering | bandpass / cavity for one service band | cavity filter |
| Powering a mast LNA | DC up the coax | bias tee; bias_tee in the sdr device blocks (config.example.yaml) |
| Judging the result | decode error rate across a gain sweep | autogain, The Hunt Part 5 |
In this post
- Why the first stage wins — the Friis rule in operator terms.
- LNA at the antenna vs at the dongle — the same part, two outcomes.
- The overload tax — when gain makes decoding worse.
- Filters: notch, bandpass, and where they go — order-of-operations rules.
- Bias tees and powering — DC up the coax without surprises.
Why the first stage wins
The Friis noise cascade says the system noise figure is dominated by the first stage, with every later stage’s contribution divided by the gain in front of it. You don’t need the formula to use the consequence:
- Loss before the first amplifier counts in full. Part 8’s 6.5 dB of RG-58 ahead of a bare dongle is 6.5 dB of system noise figure.
- Loss after ~20 dB of low-noise gain barely counts. The same 6.5 dB of coax behind a 20 dB LNA contributes a fraction of a dB to the cascade.
That asymmetry is the entire placement argument. A typical SDR-grade LNA has a noise figure under 1 dB and ~20 dB of gain; put it at the antenna and the system noise figure becomes “about 1 dB, plus a little” almost regardless of the feedline behind it. The coax run stops being a noise problem and becomes a mere plumbing problem.
LNA at the antenna vs at the dongle
Same amplifier, two different systems:
| Configuration | Approx. system NF | What actually happened |
|---|---|---|
| antenna → 40 ft RG-58 (~6.5 dB) → tuner (NF ~6 dB) | ~12.5 dB | baseline: coax + tuner both count in full |
| antenna → 40 ft RG-58 → LNA at desk → tuner | ~7.5 dB | the coax’s 6.5 dB still leads the cascade |
| antenna → LNA at mast → 40 ft RG-58 → tuner | ~1.5 dB | coax and tuner both hide behind the LNA’s gain |
The desk-mounted LNA looks busy — the waterfall brightens, dBFS rises — but most of what it amplified was a signal already degraded by the run. Six dB of the improvement you paid for never existed. The mast-mounted unit changes the system class. This is also the honest way to read Part 8’s budget: replacing cable and adding a mast LNA attack the same dB from two sides, and the LNA is usually the bigger single win — if the band lets you spend the headroom.
The overload tax
Here is the trade nobody prints on the LNA’s box: every dB of gain ahead of the tuner is a dB less headroom inside it. The LNA doesn’t know which carrier you want. It lifts the 100 kW FM broadcast transmitter at 98.7, the paging blowtorch at 152.48, and the neighbor’s cellular uplink right along with your 851 MHz control channel — and Part 4 showed what happens when the sum gets big: intermod products that look like phantom carriers, a rail-pinned ADC behind a deceptively normal FFT (the nineteen-dibits postmortem), and a decode error rate that rises with gain on the far side of Part 3’s U-curve.
The operator-level rule: an LNA is a rural instrument by default. On a quiet band it’s nearly free noise figure. In an urban RF environment it must arrive with a filter, and you should expect to reduce the tuner’s own gain after installing it — re-run the gain ladder from Part 3 and let decode quality, not dBFS, pick the new operating point. Remember the standing rule from #764: neither of those captures clipped, and the front end was still the problem — “no clipping” does not clear an overloaded or noisy chain, and a brighter waterfall proves nothing.
Filters: notch, bandpass, and where they go
A filter spends a little in-band insertion loss (typically 0.5–3 dB depending on type) to remove out-of-band power before it can do harm. The two workhorses for a trunking rig, both covered with buying picks in the SDR filters guide:
- FM broadcast notch (reference) — kills 88–108 MHz, which is the single most common overload source feeding a wideband antenna. Cheap, nearly universal benefit for anything above VHF.
- Bandpass for your service band — e.g. a 700–900 MHz pass for a trunking rig, or a proper cavity filter when one specific neighbor is the problem. Narrower is better protection and less flexibility.
Position is a decision rule, not a dogma:
| Situation | Order | Why |
|---|---|---|
| Quiet band, weak target | antenna → LNA → filter → coax | filter’s insertion loss hides behind the LNA’s gain; NF stays minimal |
| Hot band (FM/pager towers in sight) | antenna → filter → LNA → coax | the LNA itself is what overloads; it must be protected even at the cost of NF |
| Overload only at the tuner, LNA clean | antenna → LNA → coax → filter → tuner | spend the loss where it’s cheapest |
The first row is the textbook default; the second is the one that saves urban installs. If you’re unsure which regime you’re in, the test from Part 4 applies: insert 10 dB of attenuation at the tuner — if signals appear or decode improves, you’re overloaded somewhere, and the filter goes in front of whatever stage the attenuator just rescued.
Bias tees and powering
A mast-mounted LNA needs DC, and running a second cable up the mast defeats
the purpose. A bias tee injects
supply voltage onto the coax’s center conductor; the LNA end extracts it, and
the RF rides through both ways. Many SDRs can source this directly — the
device blocks in config.example.yaml expose it as bias_tee: true
(RTL-SDR Blog V3/V4 have one
built in; on sdr.soapy_remote devices it’s best-effort, driver-dependent). Two cautions: a DC short anywhere in the run (some antennas
are DC-grounded by design) will make the dongle fold back its supply — check
the antenna’s spec before enabling it; and an inline component that doesn’t
pass DC (most filters don’t) must sit on the radio side of the injection
point, not between the bias tee and the LNA.
Where this goes next
The chain from antenna to ADC is now as good as we can make it: the right antenna (Part 7), low-loss feed (Part 8), clean gain in the right order (this part). Before we spend money on a second antenna, we need the skill that turns any remaining mystery into evidence: Part 10 covers capture discipline — the formats, sidecars, and tap points that let a “sounds bad” complaint become a replayable file, which is the tracker’s most repeated request and the reason half its hard bugs ever got solved.
FAQ
Do I need an LNA at all with a modern SDR? Maybe not. If your feedline is short and low-loss and your decode is clean, an LNA buys you nothing but overload risk. The LNA’s case is strongest with long runs (it neutralizes the coax) and weak signals. Fix the feedline first — Part 8 is cheaper than any amplifier.
How much LNA gain should I buy? Enough to bury the losses behind it — 15–20 dB covers any sane feedline. More gain than that subtracts straight from overload headroom without improving noise figure, since the cascade is already dominated by the first stage.
Filter first or LNA first? Quiet band: LNA first (its gain hides the filter’s insertion loss). Hot band: filter first, because the LNA is the stage being overloaded and a fried noise budget beats a spur farm. When in doubt, run the attenuator test from Part 4 and see which stage the overload lives in.
Will GopherTrunk tell me my LNA is helping? Indirectly and honestly: sweep tuner gain and compare decode error rate before and after, the same way autogain scores gains. Expect the optimal tuner gain to drop after adding an LNA. If error rate at the new optimum didn’t improve, the LNA is amplifying a problem — or sitting in the wrong place.
Can I power a mast LNA through a splitter or a second receiver? Treat DC paths explicitly. Splitters, filters, and some antennas block or short DC, and a bias-tee feed that works through one path can vanish through another. One injection point, one extraction point, and check every inline component’s DC behavior — “LNA stopped amplifying after I added a filter” is almost always a starved bias tee.
Series navigation
Part 9 of 14 · ← Part 8: Feedline & Connectors — Where dB Go to Die · Next → Part 10: Capture Discipline — cfile, cs16, SigMF & Metadata