Linear phase filters

User avatar
UliBru
Posts: 434
Joined: 19 Oct 2019 13:58

Re: Linear phase filters

Post by UliBru » 22 Jun 2020 09:18

By CutNWindow you cut a part out of a longer pulse. Which part you have to cut depends on the position of the main pulse peak. It contains the most information about the pulse.
Now keep in mind that minphase filters have the main peak at the pulse start whereas linearphase filters have the peak in the pulse center.
So for minphase use the start position 0 whereas for linearphase you can use the "peaksymmetric" checkbox.
It may become a bit more difficult if you convolve minphase and linearphase. The result is not necessarily peaksymmetric.
Here is an example:
the result is a convolution result of 3 linearphase filters (65536 taps) and a minphase filter (65536 taps).
The overall length is 262144 taps. By the 3 linphase filters the peak is at position 3 x 32768 = 98304. You would cut a 65536 taps extract either peaksymmetric or at start position 98304 - 32768 = 65536. The additional convolution with a minphase filter does not change the start position.
So finally apply CutNWindow with 65536 taps length at start position 65536.
Stay well tuned
Uli

Moderator
Acourate system: JRiver/Roon -> AcourateConvolver -> miniDSP U-DIO8 -> TacT M/S2150 amps -> DIY horn speakers
sadhill
Posts: 19
Joined: 22 Oct 2019 10:20

Re: Linear phase filters

Post by sadhill » 22 Jun 2020 12:56

Ok, thank you.

I made some trials, quite tricky (for example, with 1 original linear filter - that is : not convoluted - and 4 convolutions to add 4 minphase filters, I get a peak at 32902 given in the cut'n window window, but I am unable to find it by calculation...)

To make it simpler, if the resulting curve is overlapping exactly the original one, does it mean it's ok ?
User avatar
UliBru
Posts: 434
Joined: 19 Oct 2019 13:58

Re: Linear phase filters

Post by UliBru » 22 Jun 2020 13:02

1 linphase and 4 minphase filters convolved...
This means that the minphase filters create a peak at position 134. Plus 32768 for linphase you get 32902.
So if you apply CutNWindow still select a start position 0.
As you cut the tail by 196608 samples you cut away information. Depending on it the resulting curve may look similar but it also can look different.
Stay well tuned
Uli

Moderator
Acourate system: JRiver/Roon -> AcourateConvolver -> miniDSP U-DIO8 -> TacT M/S2150 amps -> DIY horn speakers
sadhill
Posts: 19
Joined: 22 Oct 2019 10:20

Re: Linear phase filters

Post by sadhill » 22 Jun 2020 14:59

not so easy to understand for unskilled people, but enough to get results... thank you Uli ! Hopefully I'll become a specialist in a future life ;)
Post Reply