English is currently available.
Cellular, GSM & LTE DISCUSSION
My board has a 2G module, a small audio amplifier for voice prompts and an analog pressure sensor on the microcontroller's ADC. Whenever the module registers or sends data, the speaker produces a loud rhythmic buzz and the ADC readings jump by several counts. It happens even when the audio input is muted.
Where does this noise come from, and what are the accepted ways of keeping it out of the audio and analog parts of the design?
The buzz is the envelope of the transmitter. GSM divides time into frames of 4.615 ms containing eight slots of 577 µs, and the module transmits in one slot per frame. The RF carrier is therefore switched on and off about 217 times a second (1 / 4.615 ms = 216.7 Hz), at up to 2 W peak in the 850/900 MHz bands and 1 W at 1800/1900 MHz. That repetition rate and its harmonics are right in the audible range.
The carrier itself is far above audio, but any semiconductor junction it reaches, such as an amplifier input stage or an ESD diode, rectifies it like a crude AM detector and recovers the envelope. Muting the audio source does not help, because the pickup happens in the wiring and in the amplifier input itself.
There is a second, conducted path. Each burst pulls up to 2 A from the supply, so the module's supply rail and the shared ground carry a 217 Hz pulse train. If the amplifier or the sensor shares that ground return, a few milliohms of copper turn the current into noise: 2 A through 10 mΩ is 20 mV, which is large compared with a small audio signal or with one LSB of a 12-bit ADC at 3.3 V (0.8 mV).
Give the module its own supply branch with a large low-ESR capacitor at its pins, return its ground current directly to the supply entry point, and keep analog ground currents out of that path. A solid ground plane with thoughtful placement works better than cutting the plane into islands.
For the radiated path, stop the RF before it reaches a junction. Place small capacitors from each audio and sensor line to ground right at the amplifier or ADC pin, chosen so that their self-resonance falls near the carrier: values around 33 pF for the 900 MHz bands and around 10 pF for 1800/1900 MHz are the usual suggestion in module design guides. A ferrite bead or a small series resistor ahead of the capacitor forms a low-pass filter. Large capacitors alone do little, because at 900 MHz their parasitic inductance dominates.
Keep audio traces short, route microphone and speaker lines as differential pairs where the parts allow it, and place the antenna and its cable as far from the analog section as the board permits.
For the ADC there are firmware measures as well. Add an RC filter at the ADC input, for example 1 kΩ and 100 nF (corner near 1.6 kHz), if the sensor's output impedance and the converter's sampling requirements allow it, and average over a whole number of burst periods: 16 samples spread evenly over 4.615 ms cancels most of the periodic component. If the module only transmits when you ask it to, simply do not sample during transmissions.
To find which path dominates, run the module from a separate battery while it transmits: if the noise stays, it is radiated pickup; if it drops, it is supply and ground coupling. LTE in FDD bands does not use this 217 Hz burst pattern, so the effect is usually much weaker there, but it returns whenever a multi-mode module falls back to 2G.
Back to Cellular, GSM & LTE community
