Notch Filter Attenuation (FILT1_NOTCH_ATT)

Description

FILT1_NOTCH_ATT defines how much of the noise is removed. A value of 0.1 reduces the signal at that frequency to 10% of its original strength.

Notch Filter Center Frequency (FILT1_NOTCH_FREQ)

Description

FILT1_NOTCH_FREQ defines the "Target" of a specialized electrical filter used to clean up noisy sensor data.

A Notch Filter removes one specific frequency while leaving others untouched. This is most commonly used to eliminate the vibration frequency of a drone's motors from the accelerometer or gyroscope data, allowing for much sharper and more stable flight control.

Tuning & Behavior

• Default: 0 (Disabled).
• Usage: Find your frame's peak vibration frequency using the FFT diagnostic tool or by analyzing logs. Enter that frequency here.
• Effect: Reduces mechanical "noise" that would otherwise confuse the PID controller.
• Integration: Must be used with FILT1_NOTCH_Q to define the width of the filter.

Notch Filter Quality Factor (FILT1_NOTCH_Q)

Description

FILT1_NOTCH_Q determines how narrow the "Notch" is.

• Higher Q (e.g. 2.0): Narrow notch. Removes a very specific frequency without affecting nearby data.
• Lower Q (e.g. 0.5): Wide notch. Removes a broader range of frequencies around the center.

FlowHold Filter Frequency (FILT_HZ)

Description

This parameter sets the cutoff frequency for the low-pass filter applied to Optical Flow sensor data when using FlowHold mode. Optical flow sensors (which measure the "movement" of the ground beneath the drone) can be noisy due to surface texture, lighting, or vibration.

A lower frequency makes the drone's position hold feel "softer" and more resistant to noise, but it adds latency (delay) to the control loop. A higher frequency makes the drone react more instantly to movement but can cause "jitter" if the sensor data is noisy.

The Mathematics

The filter is a standard first-order Low Pass Filter (LPF). The relationship between the cutoff frequency ($f\_c$) and the filter's time constant ($\tau$) is:

$$ \tau = \frac{1}{2 \pi f\_c} $$

The smoothing coefficient ($\alpha$) used in each update step is:

$$ \alpha = \frac{dt}{dt + \tau} $$

Where $dt$ is the loop time (typically 0.0025s for a 400Hz control loop). At the default of 5Hz, the time constant is approximately 0.031s, providing significant smoothing for raw flow data.

The Engineer's View

In AC_PI_2D.cpp, the parameter _filt_hz is used to calculate the filter coefficient.

1. The code enforces a minimum frequency of 0.01Hz to prevent division-by-zero errors.
2. In the update() loop, the raw flow error is passed through this filter before the Proportional (P) and Integral (I) gains are applied.
3. This ensures that high-frequency noise from the sensor does not "kick" the motors, which would cause audible chirping and wasted battery energy.

Tuning & Behavior

• Default Value: 5 Hz.
• Range: 1 to 100 Hz.
• Effect of Increasing: The vehicle will respond more aggressively to small drifts. Use this if the drone feels "lazy" in FlowHold but you have a very clean, high-quality sensor (e.g., HereFlow).
• Effect of Decreasing: The vehicle will feel more stable but may drift further before correcting. Use this if you see the drone "shaking" or "twitching" in a hover.
• Dependencies: Primarily used when FLT_MODE is set to FlowHold.