Conducted noise immunity

The noise immunity is an important characteristic in the evaluation of a touch sensing system.

The standard IEC61000-4-6 details the test environment for the conducted noise.

The IEC61000-4-6 standard specifies the test procedure to evaluate the conducted noise immunity of an EUT (equipment under test).

The figure below shows how the modulated noise signals are injected into the equipment under test.

Touch sensing systems are based on capacitor variation measurements. The system must be able to detect capacitive variations as low as few picofarads on the sensor electrodes. Therefore, such systems may be sensitive to conducted noise.

In the test setup shown in Figure 1 Standard IEC61000-4-6 test setup, the injected signal simulates the noise perturbations to which a system may be exposed. By varying the frequency and the level of the injected signal, the test setup allows the characterization of situations where the touch system becomes unreliable.

The injected signal is a swept modulated noise source with a sine wave envelope as shown in the figure below.

The noise generator frequency range is swept from 150 kHz to 80 MHz. The frequency is swept incrementally. The step size must not exceed 1 % of the preceding frequency value. The signal is 80 % amplitude modulated with a 1 kHz sine wave.

The modulated noise signal amplitude may be expressed either in V_rms or V_pp.

Here is the formula to convert values from V_rms to V_pp:

$Vppvalue=Vrms\times \sqrt{2}\times 1.8\times 1.2$

The EUT noise immunity is evaluated by testing the ability of the EUT to behave according to the definition of a given class when it is submitted to a given noise level. The table below summarizes the different noise levels and classes.

The minimum frequency step recommended by the standard to sweep the injected signal from 150 kHz to 80 MHz is 1 % of the preceding frequency value. At 500 kHz this represents a 5 kHz step.

On most touch sensing systems, these steps are too large to be able to isolate the worst case situations. Some applications with narrow critical wave band can pass 3 V_rms if the critical frequency falls just in the middle between tested frequencies. The same application does not pass 1 V_rms if the user sets the test exactly on the critical frequency.

It is then important to set smaller steps around the critical frequencies. Sometime the standard bench is not able to do the appropriate step (for example 100 Hz). Test set up proposal to detect worst case proposes a method to detect the worst case.

The STM32 microcontrollers use a surface charge transfer acquisition principle, briefly described below.

The surface charge transfer acquisition principle consists in charging a sensor capacitance (C_x) and transferring a part of the accumulated charge into a sampling capacitor (C_s). This sequence is repeated until the voltage across C_s reaches a given threshold (V_IH in our case). The number of charge transfers required to reach the threshold is a direct representation of the size of the electrode capacitance.

When the sensor is touched, the sensor capacitance to the earth is increased so the C_s voltage reaches the threshold with less count and the measurement value decreases. When the measurement value falls below a defined threshold, a detection is reported.

The noise injection disturbs the measurement proportionally to its amplitude and depending on its frequency. The worst case is generally found at a noise frequency close to the charge transfer frequency (assuming no techniques are used to spread this frequency).

The figure below describes the C_s voltage behavior with and without finger on the sensor. The finger capacitance added on the sensor speeds up the C_s charge. When V_cs reach V_IH, the TSC peripheral stops charging the transfer state machine.

After this step, a discharge is done and V_cs goes from V_IH to 0 V.

The figures below describe C_s voltage behavior with and without finger on sensor during conducted noise injection.

The measure value increases or decreases with the conducted noise frequency range. Regarding the delta value, the behavior is less predictable. The TSL (touch sensing library) debounce feature helps to filter these effects (see debounce and threshold setting in Software filter (debounce)).