GFSK vs FSK: Filtering, Spectrum and Sensitivity

Key Takeaways

• The Gaussian filter smooths the data transitions, so the transmitted frequency slides between tones instead of jumping.
• Smoothing narrows occupied bandwidth and suppresses the sidelobes that leak into neighbouring channels.
• A lower bandwidth-time product narrows the spectrum further but increases intersymbol interference.
• Unfiltered FSK keeps sharper symbols, which can help sensitivity and frequency-error tolerance at a given deviation.
• The choice is a spectrum-versus-receiver trade, decided by the channel plan and the data rate you need.

1. What Does the Gaussian Filter Change in GFSK?

In GFSK, a Gaussian low-pass filter shapes the baseband data before it reaches the modulator, so the transmitted frequency changes smoothly rather than instantly.

FSK and GFSK map the same data onto the same two tones. What changes is the shape of the transition between them, and that shape drives occupied bandwidth, adjacent-channel power and receiver margin.

Core characteristics

• The filter sits in the baseband path. It acts on the symbol stream, not the carrier, so the same shape applies at any centre frequency.
• Smooth transitions narrow the spectrum. A gradual frequency change spreads less energy, so occupied bandwidth falls at a given symbol rate.
• Sidelobes decay faster with offset. Rounded transitions push less energy towards neighbouring channels as you move from the carrier.
• The receiver pays for the narrowing. Adjacent symbols overlap more, so the eye closes and more signal-to-noise is needed.

2. How the Gaussian Filter Works Before the Modulator

A GFSK transmitter is an FSK transmitter with one extra block in front of the modulator, and its consequences follow from that block.

1. The data passes through a Gaussian low-pass filter. Each sharp rectangular pulse becomes a smooth, bell-shaped one. It is a baseband choice made at design time, and it changes the spectrum, not the data rate.
2. The filtered waveform drives the modulator. The carrier slides from one tone towards the other instead of stepping, so the transmitter never presents an abrupt frequency step to the power amplifier.
3. Smoothing narrows the spectrum and suppresses sidelobes. A hard transition is a broadband event, and rounding it removes the high-frequency content that spreads the spectrum. Less energy lands in the neighbouring channel.
4. The BT product sets how much smoothing you get. A lower product means a narrower filter and a tighter spectrum; a higher product approaches the unfiltered case, and narrowing reshapes the symbol.
5. Narrowing spreads each symbol into its neighbours. A symbol's energy leaks into adjacent periods, so intersymbol interference grows, the eye closes, and more signal-to-noise is needed for the same error rate.

3. What Is Unfiltered FSK?

Unfiltered FSK switches the instantaneous frequency abruptly between two tones at each symbol boundary, with no baseband shaping in front of the modulator.

Remove the Gaussian filter and you return to the original implementation: two tones, switched cleanly, with the receiver deciding tone by tone. That different set of priorities favours the receiver.

Core characteristics

• The transition is a step, not a slide. Each symbol starts and ends on a well-defined tone, so the receiver sees a clean change per bit period.
• Abrupt edges spread energy into sidelobes. A fast frequency change has wide spectral content, so energy appears far from the carrier and decays slowly.
• Occupied bandwidth is wider at the same symbol rate. Carrying the same rate needs more spectrum, which limits how tightly channels pack.
• Symbols stay separable. With no smearing across symbol periods the eye stays open, which can improve sensitivity and tolerance to frequency error.

4. GFSK vs FSK: What Is the Difference?

Both modes carry the same information at the same symbol rate. The difference is where the shaping happens and who pays: the filter trades spectral purity for receiver margin.

The decision rule: choose GFSK when the signal must fit a tight channel or a regulatory mask; choose unfiltered FSK when the band is open and you want the last few dB of margin at a fixed deviation.

5. Key Parameters in a GFSK or FSK Link

The parameters below decide whether a filtered or unfiltered link works, and each shows up as a different symptom when it is wrong.

• Deviation: sets tone spacing and occupied spectrum. Set it so the tones stay resolvable at your receiver bandwidth. Wrong: too little merges the tones; too much breaches the mask.
• Data rate against occupied bandwidth: both rise together. Pick the lowest rate with payload headroom. Wrong: too high a rate spreads beyond the channel and fails at the band edges.
• Bandwidth-time product: sets how hard the filter smooths. Smaller product in congested bands, larger where the band is open. Wrong: too small a product closes the eye.
• Receiver bandwidth: must match deviation and rate. Wide enough to pass the deviation, narrow enough to reject noise. Wrong: too narrow clips the tones; too wide costs sensitivity.
• Crystal accuracy and drift: sets the carrier, and both ends must stay inside the receiver passband. Use a stable, low-drift oscillator. Wrong: links that work warm fail at temperature extremes.
• Adjacent-channel power: the energy that lands next door. Check it against the spectrum mask your regulator applies. Wrong: a clean bench link degrades when a second transmitter starts.
• Air data rate against range: a higher air rate needs more bandwidth and shortens reach at the same power. Wrong: links pass at short range and drop packets at the far end.
• Preamble and sync word length: how the receiver detects frames and settles timing. Lengthen them for longer or heavily filtered links. Wrong: missed frames despite an adequate link budget.
• Antenna and receive filter bandwidth: both must pass the band you use, with a short feed. Wrong: range far below the datasheet because the antenna detunes.

6. When to Use GFSK and When Not To

Good fit for GFSK

• Congested sub-GHz bands, where adjacent-channel power and channel spacing set the limit.
• Spectrum masks that a wide unfiltered signal would breach at the rate you need.
• Dense sites with many transmitters, where coexistence matters as much as range.

Good fit for unfiltered FSK

• Open spectrum with generous channel spacing, where extra bandwidth costs nothing.
• Links where sensitivity and frequency-error tolerance matter more than spectrum.
• Simple transmitter paths with no baseband filter to configure.

Poor fit for GFSK

• Very high data rates, where smoothing limits how cleanly symbols separate.
• Links with no receiver margin left to give away.
• Designs that cannot absorb validating a BT product.

Poor fit for unfiltered FSK

• Crowded channel plans with a neighbouring transmitter one channel away.
• Bands governed by a strict spectrum mask at the data rate you need.
• Busy long-range sites, where interference rather than sensitivity sets the limit.

7. Real-World Applications

In sub-GHz industrial links the modulation choice is made once, at module level, then hidden behind a configuration parameter.

For filtered operation, several Ebyte modules use GFSK. The E70-433T14S2 is a UART transceiver on the TI CC1310F64RSM with hardware forward error correction, covering 431 to 446 MHz (default 433 MHz) at 14 dBm. The E70-900T14S2 spans 861 to 922.5 MHz with a 1.5 km figure and industrial-grade low-temperature-drift crystal oscillators. The E49-400T20D is a 433 MHz DIP transceiver at 20 dBm.

Where a design must switch between wide-area and filtered narrowband links, the E22-400M22S and E22-900M22S use multiple modulation methods including LoRa and GFSK over SPI.

In the 2.4 GHz band GFSK is the native Bluetooth modulation. The E104-BT53A3 is a small SMD Bluetooth 5.2 module on the Silicon Labs EFR32BG22, covering 2402 to 2480 MHz with a 120 m figure.

Measured range depends on the antenna and its feed as much as on the modulation. The E23-433M13S uses an IPX interface with an SX1212 at 433 MHz and 13 dBm. Keeping the feed short over the last few centimetres accounts for more field failures than the mode choice.

8. FAQ: GFSK and FSK in Practice

Q1: What is the difference between GFSK and FSK?

GFSK is FSK with a Gaussian low-pass filter applied to the data before modulation, so the transmitted frequency slides between tones instead of jumping. The filter narrows the occupied spectrum and reduces adjacent-channel power, but it also spreads each symbol into its neighbours and can cost sensitivity.

Q2: What does the BT product control in GFSK?

The bandwidth-time product sets how strongly the Gaussian filter smooths the data, by comparing filter bandwidth with symbol rate. A lower BT product gives a narrower spectrum and lower adjacent-channel power but more intersymbol interference; a higher one approaches unfiltered FSK and recovers margin.

Q3: Does GFSK reduce range compared with FSK?

Not directly, but it can reduce the margin you have left. GFSK narrows the spectrum and lowers adjacent-channel power, and the cost is intersymbol interference, which closes the eye and raises the signal-to-noise needed for the same error rate. In a quiet band, unfiltered FSK can reach further at a fixed deviation.

Q4: Why does my GFSK link lose packets when a nearby transmitter runs?

Another transmitter's energy is landing inside your channel, or yours is landing inside theirs. Work through these checks:

• Packet loss starts when the other transmitter keys -> check the frequency offset between the two carriers.
• Loss appears at the band edge -> check occupied bandwidth and deviation.
• Loss at close range but not at distance -> check the receiver front end for blocking.
• Loss that follows temperature -> check crystal drift between the two carriers.

Q5: Is GFSK always better than FSK?

No. GFSK buys spectral efficiency at the cost of receiver margin, so it is the better choice when the spectrum is the binding constraint. If the band is open and the receiver is the weak point, unfiltered FSK can be stronger.

Q6: How do I choose a BT product for a sub-GHz link?

Start from the channel plan and the spectrum mask, not from a single recommended value. Pick the largest BT product that still keeps occupied bandwidth and adjacent-channel power inside the mask, then verify sensitivity on the bench.

9. Pre-Flight Checks Before You Choose a Modulation Mode

1. Map the channel plan first. Note the channel spacing, the spectrum mask your regulator applies, and how many transmitters share the band. Choose GFSK when those numbers are tight.
2. Set deviation and receiver bandwidth together. Confirm the tones stay separable inside the demodulator bandwidth, then check occupied bandwidth against the channel you have.
3. Fix the BT product by measurement. Test occupied bandwidth, adjacent-channel power and sensitivity, and keep the largest value that passes the mask.
4. Budget the margin you give away. Compare sensitivity and packet error rate for both modes at the same deviation.
5. Recheck the antenna and crystal. Range figures assume a matched antenna and a stable reference.