devices.xylo.syns63300.imuif.BandPassFilter

class devices.xylo.syns63300.imuif.BandPassFilter(B_b: int = 6, B_wf: int = 8, B_af: int = 9, a1: int = -36565, a2: int = 31754, scale_out: float | None = None)[source]

Bases: object

Class that instantiates a single quantised band-pass filter, as implemented on Xylo IMU hardware

This class will design the best filter that meets a pass-band specification, using the class method from_specification(). Alternatively, you can specify the a1 and a2 taps directly as a signed integer representation, as well as specifying the number of bits needed for specifying the filter.

Attributes overview

B_af

Bits needed for encoding the fractional parts of taps

B_b

Bits needed for scaling b0

B_wf

Bits needed for fractional part of the filter output

a1

Integer representation of a1 tap

a2

Integer representation of a2 tap

scale_out

A virtual scaling factor that is applied to the output of the filter, NOT IMPLEMENTED ON HARDWARE! That shows the surplus scaling needed in the output (accepted range is [0.5, 1.0])

Methods overview

__init__([B_b,Β B_wf,Β B_af,Β a1,Β a2,Β scale_out])

compute_AR(signal)

Compute the AR part of the filter in the block-diagram with the given parameters.

compute_MA(signal)

Compute the MA part of the filter in the block-diagram representation.

from_specification(low_cut_off,Β high_cut_off)

Create a filter with the given upper and lower cut-off frequencies.
B_af: int = 9

Bits needed for encoding the fractional parts of taps

B_b: int = 6

Bits needed for scaling b0

B_wf: int = 8

Bits needed for fractional part of the filter output

__init__(B_b: int = 6, B_wf: int = 8, B_af: int = 9, a1: int = -36565, a2: int = 31754, scale_out: float | None = None) → None
a1: int = -36565

Integer representation of a1 tap

a2: int = 31754

Integer representation of a2 tap

compute_AR(signal: ndarray) → ndarray[source]

Compute the AR part of the filter in the block-diagram with the given parameters.

Parameters:

signal (np.ndarray) – the quantized input signal in python-object integer format.

Raises:

OverflowError – if any overflow happens during the filter computation.

Returns:

the output signal of the AR filter.

Return type:

np.ndarray

compute_MA(signal: ndarray) → ndarray[source]

Compute the MA part of the filter in the block-diagram representation.

Parameters:

signal (np.ndarray) – input signal (in this case output of AR part) of datatype pyton.object.

Raises:

OverflowError – if any overflow happens during the filter computation.

Returns:

quantized filtered output signal.

Return type:

np.ndarray

classmethod from_specification(low_cut_off: float, high_cut_off: float, fs: float = 200) → BandPassFilter[source]

Create a filter with the given upper and lower cut-off frequencies. Note that the hardware filter WOULD NOT BE EXACTLY THE SAME as the one specified here. This script finds the closest one possible

Parameters:

  • low_cut_off (float) – The low cut-off frequency of the band-pass filter.

  • high_cut_off (float) – The high cut-off frequency of the band-pass filter.

  • fs (float) – The clock rate of the chip running the filters (in Hz). Defaults to 200.

Returns:

the filter with the given cut-off frequencies.

Return type:

BandPassFilter

scale_out: float | None = None

A virtual scaling factor that is applied to the output of the filter, NOT IMPLEMENTED ON HARDWARE! That shows the surplus scaling needed in the output (accepted range is [0.5, 1.0])