Receive Filter Bandwidth: Roofing Filters, DSP and Choosing Widths

If you could change only one setting on your transceiver to hear more stations, the best candidate would be the receive filter bandwidth. It decides how much of the band reaches your ears at any moment. Too wide and you hear every neighbour, plus more noise than you need. Too narrow and signals sound thin, or you lose the station you wanted because it is a little off frequency. This guide explains how to choose a sensible width for CW, SSB and digital modes, what the roofing filter and the DSP filter each contribute, and how shift, notch and similar controls let you carve a signal out of a crowded band without buying anything new.

Why Bandwidth Matters

Every receiver listens through a window. Inside the window you hear the signal, the noise and any other transmission that happens to be there. Outside it, the filter removes the sound. For random noise, the power that gets through is proportional to the bandwidth, so halving the bandwidth cuts the noise by about 3 dB. Moving from a 2.4 kHz SSB filter to a 500 Hz CW filter reduces the noise power by roughly 7 dB, and that improvement shows up as a better signal-to-noise ratio for the same CW signal. The general relationship between bandwidth and what passes a filter is described in the article on bandwidth in signal processing.

That is the first reason to narrow the filter: you hear the signal with less noise. The second reason is selectivity. When a strong station sits only a few hundred hertz away, a narrower filter attenuates it, and the weak signal becomes readable. The third reason is that the filter protects the later stages of the receiver from strong out-of-passband signals, which is a different job that we will reach in the roofing filter section.

The limit is the signal itself. A voice needs about 2 to 3 kHz to be understood and natural. A CW signal needs only a few hundred hertz, depending on speed and how softly the transmitter shapes the dots and dashes. A weak-signal digital mode may need only a few tens of hertz. The filter should be as narrow as the signal permits and no narrower.

Two Kinds of Filters in One Radio

In a superheterodyne receiver the signal passes through a chain of filters at different frequencies, and the amateur-radio conversation often mixes up the types. It helps to separate them.

The Roofing Filter

The roofing filter is a crystal or ceramic filter that sits early in the receiver, after the first mixer. Its job is to limit the bandwidth of the first intermediate frequency stage so that strong signals outside the channel cannot overload the amplifiers and mixers that follow. It does not set the final bandwidth you hear. The narrower the roofing filter, the fewer strong signals reach the later stages, and the better the receiver handles a crowded band. Broad roofing filters, of several kilohertz, are enough for general use, while narrow roofing filters of a few hundred hertz to the SSB width are used by operators who work contests or crowded bands.

Many modern radios offer a choice of roofing filters or select one automatically based on the mode and the bandwidth you chose. If your radio lets you pick, a narrower roofing filter will usually help on the busy phone and CW segments, at the cost of losing the wide view you need for scanning the band. Check the manual for how your model does it, as the options and their names differ between manufacturers. For the background on why this matters for strong-signal performance, see our guide to receiver dynamic range.

The DSP or Final Filter

The final bandwidth you hear is set by a later filter, which in modern radios is a digital filter. A DSP filter can have a very steep skirt, a precise width and a chosen shape, and it can be adjusted in small steps. It cannot protect the front end from overload, because by the time the signal reaches it the overload would already have happened. That is the key distinction: the roofing filter protects, and the DSP filter shapes.

For software-defined receivers the same principle applies in a different form. The converter sees a wide slice of spectrum, and the strong signals inside that slice limit what can be heard, whatever the digital filter does later. Our comparison of direct sampling SDR and superheterodyne receivers explains how the two designs treat strong signals.

Choosing a Width for Each Mode

The table below lists typical starting points. They are what many operators use, not rules, and you should always adjust to the conditions. The single best test is to narrow the filter until the signal begins to lose quality, and then open it one step.

SSB

A single-sideband voice signal is described in the article on single-sideband modulation and normally fits in about 2.4 to 3 kHz. A filter at 2.4 kHz is the common default. Narrowing to about 1.8 to 2.1 kHz in a pileup removes some of the highs and the lows, which makes the voice sound thinner but often makes it easier to understand through a crowded band. Opening to 3 kHz or more brings back the natural tone for friendly conversations on a quiet band. If you notice that your transmitted voice sounds too boomy or too sharp, our guide to SSB audio settings covers the transmit side of the same question.

CW

Many CW operators settle at 400 to 500 Hz for general use and go narrower for weak signals or contests. A very narrow filter has a price. Narrow filters ring, which means they continue to sound for a moment after a dit ends, and at high keying speeds the ringing smears the elements together. A filter that is too narrow also makes it hard to tune to a station, because a small error takes the signal out of the passband. The practical limit is a matter of skill and speed, and you will find it quickly by experiment. If you want to line up your receive frequency exactly, the technique is described in our article about zero beating a CW signal.

Digital Modes

For weak-signal digital modes that decode in software, the radio filter is usually left wide, at 2.4 to 3 kHz, because the decoder looks across the whole audio passband and wants a clean, flat response. A narrow filter in the radio can cut off the signals you want to decode. The decoder’s own selectivity does the narrow work. For narrowband modes that use a single signal, such as RTTY with a hardware decoder, a narrower filter helps.

Shift, Passband Tuning and the Notch

Bandwidth is only one control. Several others let you move or cut the filter without moving your frequency.

• IF shift or passband tuning. Moves the filter window up or down relative to the signal, so you can slide an interfering station out of the edge of the passband while the wanted signal stays in. This is the most useful control on a crowded SSB band and is also helpful on CW.
• Width control. In many radios the width and shift can be combined to move the upper and lower edges separately. This is called twin passband tuning and gives you a very flexible way to attack an interferer on one side.
• Notch filter. A narrow rejection filter that removes a single carrier or tone, such as a heterodyne from a nearby carrier. An automatic notch tracks the tone, and a manual one lets you place it exactly.
• Audio peak filter. Boosts a narrow band around the CW pitch in the audio. It does not remove the neighbours as a true IF filter does, but it can lift a weak signal from the noise.
• Noise reduction. Works on the audio. Do not confuse it with filtering. Our article on the noise blanker and noise reduction explains what each does.

Setting the CW Pitch

The CW pitch, the tone you hear, and the filter centre should match. If the sidetone is 600 Hz and the filter is centred on 700 Hz, the radio is not at its best. Pick a pitch that is comfortable to your ears, often between 500 and 700 Hz, then set the filter centre and the transmit offset to agree with it. The match matters because a narrow filter will cut a signal that is tuned a little off.

A Method for Pulling a Weak Signal From a Crowded Band

1. Choose the mode and the typical width from the table above.
2. Tune across the signal until you hear it at the pitch you expect.
3. Narrow the filter one step at a time. Stop when the signal quality starts to suffer or when the interference is gone.
4. If a strong signal remains on one side, use IF shift or passband tuning to move it out of the filter edge.
5. If there is a steady tone, use the notch filter.
6. If the band is very loud, switch on the attenuator and select a narrower roofing filter if the radio offers one.
7. Reset to a general-purpose width when you move on, so that you do not miss callers who are a few hundred hertz away.

That last step is important for pileups. A narrow filter makes it easy to miss someone who calls slightly off your frequency, and on a split pileup the DX station may be answering callers who are scattered over several kilohertz. The tactics in our guide to working a pileup include widening the filter when you are searching and narrowing it when you are locked on.

Common Mistakes

• Starting too narrow. A very narrow filter hides the band and makes tuning hard. Begin at the usual width and narrow it when you have found the signal.
• Using the radio filter for digital modes. Most decoders prefer a wide, flat passband. Let the software select the signal.
• Believing the number on the menu. The nominal width is a point on the response curve, usually at the 6 dB level, and the real rejection depends on the skirts.
• Forgetting the roofing filter. A narrow DSP filter cannot cure overload from strong signals that are far away. Choose the roofing filter and the attenuator as well.
• Setting the pitch and the filter independently. Match the CW pitch, the filter centre and the transmit offset.

The Bottom Line

The receive filter bandwidth is the simplest tool for improving what you hear. Narrower filters cut noise and neighbours, but they cannot be narrower than the signal needs. Use roughly 2.4 kHz for SSB, 300 to 500 Hz for CW, and a wide, flat passband for software-decoded digital modes. Remember that the roofing filter protects the receiver from strong signals while the DSP filter shapes the sound, and that shift, notch and attenuator controls add flexibility. Test each change on a live band, return to a general width when you finish, and you will find that the same radio hears more of the band with only a few settings changed.

Frequently Asked Questions

What bandwidth should I use for SSB?

About 2.4 kHz is a common default. Narrow to 1.8 to 2.1 kHz in a crowded band to reduce neighbours, and widen to 3 kHz on a quiet band for more natural speech.

What is a good CW filter width?

Many operators use 300 to 500 Hz for general work and go as narrow as 100 to 250 Hz in contests or for weak signals. Very narrow filters ring and make tuning harder, so go narrower only as far as your skill allows.

What does the roofing filter do?

It limits the bandwidth in an early stage of the receiver so that strong signals outside the channel do not overload the amplifiers and mixers that follow. It protects the receiver, and the later filter sets the final audio bandwidth.

Should I narrow the filter for FT8?

Usually not. The decoder looks across the audio passband and handles many signals at once, so a wide flat radio filter of 2.4 to 3 kHz is the usual choice. Use the software to pick the signals you want.

What is the difference between IF shift and bandwidth?

Bandwidth sets how wide the window is. IF shift or passband tuning moves the window up or down, so you can push an interfering signal out of one edge while keeping the wanted signal inside.

Why does a very narrow filter sound strange?

A narrow filter rings, continuing to sound for an instant after a signal stops, and it can also remove the harmonics that give speech or CW its character. Open it by one step if the signal sounds smeared or tinny.