Parameters used: |
(optional: ) |

Supports: | LP |

Pins used: | `IN, OUT, GROUND` |

- It's made up of ½ octave spaced
*RC*cells, so it's an approximation.

sets the maximum number of*N**RC*cells, default

.*40*- Since the
*RC*cells are fixed,

determines the upper limit of the frequency/phase.*N*

smoothens or sharpens the phase response for a better match with the amplitude or the phase around*sigma*

; the default value of*BWp*

is a compromise.*0.5*- For an integrator-like response with

@*0dB*

, set*fp*

and*BWp=fp/100*

.*G=10⋅G*

Parameters used: |
(optional: ) |

Supports: | AP, LP, HP |

Pins used: | `IN, OUT, GROUND` |

- No automatic order calculation, therefore

needs to be positive definite.*N* - The terms for LP are calculated using the recursion formula and for HP they are simply reversed, therefore HP can't be constant delay.
- Frequency scaling is a polynomial approximation, thanks to zunzun. For example: sweeping

from*N*

to*1*

with*32*

will result in a difference error of*Asc=3*

between the lowest and the highest trace and*309.33mdB*

between two adjacent traces, @*~12.5mdB*

; the errors tend to be proportional with*Asc*

.*Asc* - In AP,

can only have two values:*Asc*

⇒*Asc=0*

constant phase delay for all orders with the following exceptions:*180*^{o}

with*N=2*

,*167.103*^{o}

with*N=3*

,*178.862*^{o}

with*N=4*

and*179.95*^{o}

with*N=5*

,*179.999*^{o}
or

⇒*Asc>0*

phase shift.*N⋅90*^{o}

- There may be strange distortions in the stop-band in
`.AC`

for higher orders, that's LTspice's engine doing its best to display very high values from a very simple circuit,`.TRAN`

is unaffected. Using the alternate solver or lowering

could help dampen the errors.*Rpar*

Parameters used: |
(optional: ) |

Supports: | AP, LP, HP, BP, BS |

Pins used: | `IN, OUT, GROUND` (optional: `0.1, 0.2, 0.3` ) |

Parameters used: |
(optional: ) |

Supports: | AP, LP, HP, BP, BS |

Pins used: | `IN, OUT, GROUND` (optional: `0.1, 0.2, 0.3` ) |

- If

then frequency scaling will occur @*Asc≤Ap*

(depending on the order and the value of*Ap*

).*nT*

Parameters used: |
(optional: ) |

Supports: | LP, HP, BP, BS |

Pins used: | `IN, OUT, GROUND` (optional: `0.1, 0.2, 0.3, 0.4` ) |

- If

,*Asc≥As*

will be considered.*As*

Parameters used: |
(optional: ) |

Supports: | LP, HP, BP, BS |

Pins used: | `IN, OUT, GROUND` (optional: `0.1, 0.2, 0.3, 0.4` ) |

- Accuracy is usually 3, 4 decimals or better if

, but can drop to ~2~3 or less if using extreme values for attenuations. E.g.:*N=0*

(32nd order),*N=0*

⇒ a droop of about*fc=0 BWp=1 BWs=1.1 Ap=0.001 As=260*

towards the end of the pass-band. The errors towards the other end (*0.2mdB*

) are a bit greater, but still not grater than*Ap>2.5, As<25*

.*0.1%* - While the above is possible, imposing

for the same conditions will show a real response but wrong, with many hundreds of*N=32*

attenuation (and there may be other cases like this); it's because the elliptic nome*dB*

will get too small values for the matrix solver. Choosing the alternate solver may help a bit, but not much. In short, try to keep*q1*

when*As<120*

, overly exaggerated values will not be without consequences.*N>0*

__Warning!__Enabling the test pins will slow down the simulation considerably. If only the coefficients are needed, a trick would be to use the total time(and thus,

) with a value much less than*sim*

, e.g. for*ω*_{c}

⇒ the simulation time could be*fc=0 BWp=1 BWs=2*`.TRAN 1m`

⇒

(can be tested on the same schematic)*sim=1m*

©Vlad, 2008 - 2015