Wideband HF Receiver Headroom: Filtering, Gain and Linearity

Preselection helps only when it attenuates the offending signals before the stage they overload. Input attenuation, lower or redistributed gain, a more linear amplifier or mixer, a higher-full-scale converter, a notch, spatial rejection and improved common-mode control can also create usable margin. “Headroom” is not one receiver specification; it must be tied to a reference plane and a failure criterion.

Define the Limit Before Calling It Headroom

“Headroom” is sometimes pictured as the span between sensitivity and an overload point. That can be a useful informal picture, but sensitivity, compression, blocking and intermodulation are different tests. Subtracting two unrelated headline numbers does not create a universal dynamic-range specification.

ITU-R SM.1837-1 defines a reproducible IP3 procedure for monitoring receivers from 9 kHz to 30 MHz and above. ITU-R SM.575-3 treats strong-signal protection as a system problem involving receiver IP3 and sensitivity, bandwidth and frequency of interferers, antenna gain, feeder loss and external noise. Rohde & Schwarz's SDR measurement note demonstrates a blocking test by combining a wanted waveform and swept interferer, then recording the blocker level at a defined degradation. These are measured criteria—not synonyms.

What a Preselector Actually Changes

Suppose the input filter has passband insertion loss L_p at the wanted frequency and attenuation A_b at a blocker. The useful blocker selectivity is the difference between those two measured transfers, not the filter's nominal Q or marketing bandwidth:

If that blocker dominates compression at a later stage, its margin can improve by approximately its actual added attenuation. If two blockers produce third-order intermodulation, the result depends on attenuation at both tone frequencies and the complete cascade. Analog Devices' selectivity treatment explicitly defines rejection relative to passband insertion loss and incorporates frequency-dependent selectivity between nonlinear stages.

Preselection does not help a blocker that remains in the passband. It cannot undo distortion created in an active antenna, preamplifier, protection diode or switch ahead of the filter. It also does not cure reciprocal mixing from local-oscillator phase noise, ADC clock-jitter limits, a spurious response or common-mode pickup that bypasses the intended port.

Q Gives One Width, Not the Rejection Curve

For a simple resonator or a filter response for which loaded Q is defined from its 3 dB points, the narrowband approximation is:

At 3.5 MHz, the loaded-Q arithmetic gives the following 3 dB bandwidths. Operating-mode labels and out-of-band rejection do not follow from Q alone.

Filter order, coupling, resonator unloaded Q, source and load impedance, tuning, topology and construction determine insertion loss, skirt shape and stop-band response. For an ideal single-resonator band-pass model with loaded Q = 200 at 3.5 MHz, ±20 kHz is only about 7.9 dB down—not “blocked.” A multi-pole filter may be much steeper, while parasitic coupling can make a real enclosure worse. Measure S21 over the entire spectrum that contains credible blockers.

Why Q = 10 Does Not Guarantee 10–20 dB

A 10–20 dB improvement cannot be predicted from Q = 10 alone. Such a result requires a defined metric, blocker frequencies, measured filter response and the identity of the stage that reaches its limit.

A defensible calculation begins with the installed input spectrum and measured filter S-parameters. For each signal, transfer its power to every nonlinear stage. For two equal tones in the small-signal region, a third-order product changes at 3 dB per 1 dB change of equal tone level; for unequal attenuation the relevant cubic term scales with the particular tone combination. Compression margin, a blocking threshold and an ADC's composite peak margin follow different relationships. Once compression begins, the intercept-line extrapolation is no longer the model to trust.

A Q = 10 tuned circuit can be highly useful against distant medium-wave or upper-HF broadcasters, yet almost useless against a strong station close to the wanted frequency. Conversely, a relatively broad high-pass or notch filter can outperform a narrow preselector when one known broadcast service is the problem.

Insertion Loss Buys Rejection at a Noise Cost

A passive matched loss at the standard reference temperature has noise factor equal to its linear loss. Placed ahead of the first gain stage, a 2 dB passband loss therefore adds approximately 2 dB to receiver noise figure before later-stage terms are considered. The Friis cascade relation for matched stages is:

Use linear noise factors and available gains, with mixer image and conversion conventions handled correctly. On much of HF, external noise may exceed receiver-added noise enough that some front-end loss costs little installed SNR; at a quiet site, high frequency, lossy antenna or narrow measurement bandwidth, the same loss can matter. Measure the delivered noise, not just a generic environmental curve.

Attenuation and Gain Distribution Are Real Headroom Controls

A matched attenuator ahead of a downstream nonlinear chain reduces wanted signal, external noise and blockers together. Ideally, it improves the input-referred compression and intercept margin of that downstream chain by the attenuation value while degrading its receiver noise figure by the same amount. This is a trade, not a free improvement. When the system is strongly external-noise limited, attenuation can improve reception of strong-signal environments before receiver noise becomes dominant.

AGC helps only if the controlled gain lies before the stage at risk and responds appropriately. Gain reduction after an overloaded LNA, mixer or ADC cannot repair the damage. An AGC driven by a large blocker may also reduce the wanted signal and appear as desensitisation.

A more linear stage is also a valid remedy. Raising IIP2, IIP3 or P1dB, bypassing an unnecessary amplifier, reducing gain, redistributing gain, increasing supply/headroom within device ratings or using a more suitable mixer can move the failure threshold. “Class A” alone proves none of these results: the complete circuit must be measured at the relevant frequencies, levels and impedances.

Analog Devices' wideband receiver example makes the engineering trade explicit: a high-sensitivity path favours noise figure, while a lower-gain bypass path trades sensitivity for substantially higher IP2, IP3 and input compression. The same article notes that gain beyond what is needed to establish cascaded noise performance generally hurts dynamic range.

ADC Headroom Is a Composite-Waveform Budget

An ADC input range is a voltage specification. Converting it to dBm requires the actual input impedance and waveform convention. dBFS is referenced to converter full scale, not to 1 mW, and several carriers plus noise and impulses share the same instantaneous range.

Analog Devices MT-006 defines full-scale input power from a sine wave that exactly fills the ADC input range and requires noise to be evaluated over a stated filter noise bandwidth. IEEE 1241-2023 is the current terminology and test-method standard for ADCs. A practical receiver budget must include:

• the vector sum and crest factor of simultaneous signals rather than only the largest carrier;
• analog gain and attenuation in the active state;
• converter full-scale range, input network and overload recovery;
• noise density, SNR, SINAD and SFDR at the relevant input frequency and sample rate; and
• clock-jitter and alias responses across the analog input bandwidth.

Digital channel filtering and decimation can reduce in-band noise after conversion, but they cannot recover samples already clipped or remove intermodulation generated ahead of the converter. Analog preselection can protect the ADC; post-ADC selectivity cannot.

A Resonant Loop Is Part of the Transfer Function

A tuned receiving loop can provide useful passive selectivity, but its installed field-to-terminal transfer is not determined by isolated loop Q. Coupling, matching, tuning capacitor loss, enclosure, feedline loading and receiver impedance shape the loaded response. The loop also has a radiation pattern: spatial nulls may reject a blocker better than frequency selectivity, or installation asymmetry may fill those nulls.

Common-mode current on the feedline, power cable or control lead can create a second antenna path around the intended resonator. A choke may help when its installed common-mode impedance and placement actually reduce that path. Verify the result by comparing calibrated terminal transfer and common-mode current—not by inferring it from SWR or resonant Q.

A Receiver Headroom Test That Finds the Right Fix

1. Draw the signal chain. Mark antenna, protection, switches, preselector, LNA, mixer, IF stages, AGC detector and ADC; identify each calibration plane.
2. Survey the installed spectrum. Record signal levels, bandwidth, detector, antenna state and time statistics without allowing the survey receiver itself to overload.
3. Define failure. Choose P1dB, wanted-signal SINAD/decoder loss, an IM-product limit, ADC peak margin or another explicit criterion.
4. Measure filter transfer. Record passband loss, return loss and wide-span rejection with the actual terminations and tuning states.
5. Build a cascade budget. Propagate wanted signals, blockers and noise through gain/loss; calculate IP2/IP3 only within their valid small-signal model and refer every value to a named plane.
6. Run single-tone blocking. Combine a wanted signal and swept blocker, then record degradation versus blocker offset and level.
7. Run two-tone and composite tests. Include equal and unequal tones, realistic occupied spectra and converter crest factor.
8. Compare remedies. Test filter, notch, attenuation, gain bypass, a higher-linearity stage, antenna orientation and common-mode treatment separately and in useful combinations.
9. Repeat across state. Cover bands, gain and AGC modes, supply, temperature and source/load mismatch; record uncertainty.

Bottom Line

A preselector placed before the offending nonlinear stage can be one of the most effective ways to protect a wideband HF receiver from out-of-band blockers. Adding gain without a cascade budget is often counterproductive.

Narrowing is neither universally required nor always the right remedy. A filter cannot remove an in-band blocker, undo earlier distortion or create converter range after sampling. Q alone does not specify rejection, and no universal 10–20 dB benefit follows from Q = 10.

Define the failure and reference plane, measure the blockers and transfer functions, then allocate gain, loss, selectivity, linearity and converter margin. The best receiver is not simply the narrowest one; it is the chain that preserves the wanted signal while preventing every earlier stage from becoming the interferer.

Questions, antenna-factor records or height trials to share? Contact RF.Guru.