Fix/biquad

[grrrr]

Hi,
here are a few fixes to AudioFilterBiquad_F32.

The most important is the change that begin() is now called on setCoefficients().
The reason is that setCoefficients enables doBiquad – but in update(), when doBiquad is true but arm_biquad_cascade_df1_init_f32 has not been called, this is an undefined state, causing teensy to reboot.

[boblark]

This was Chip's creation, so, Chip if you have thoughts, jump in. Otherwise, in looking at the issue, first, it needs a fix. Calling the filter without initialization doesn't work.

But, standing back, it looks like begin() and the unlisted, but available end() are intended to allow the group of biquads to be used or bypassed, leaving their definitions in place. That would seem to be the role of doBiquad==true or false.

• At begin() the filters are no longer bypassed in update() by setting doBiquad=true. As you point out, this is OK so long as the filters are initialized and so this needs to be enforced by a private flag. Then future calls to begin() or end() just turn the already-defined filter on or bypassed. So, what do you think about:

• Only changing doBiquad in begin() and end().

• Adding a filterDefined=true or false, initialized to false, that is set to true when begin() is called but with no way to reset it to false.

• Make update() do a simple bypass if either doBiquad==false or filterDefined==false.

I think that covers the issue of running filters uninitialized. And I don't think it breaks existing code.

And thanks for the filter counting changes.

Thoughts? Bob