I'm going over the FX book and learning SO MUCH ;-). Simply put this book is a gem for me. Going deeper into the DSP world, does raise many questions...For instance:
On page 283 (chapter 8, FIR filters, Moving Average) it says:
"You can see that this is a pretty poor LPF filter compared to some of the optimal filter designs, but these filters can be very useful, for example to smooth the response of an audio detector output or to insert in the feedback path of a delay line module to gently roll off high frequencies on each pass through the delay"
And I wonder:
An envelope Follower can be made with different filters, such as 1st , 2nd or higher order IIR filters. Such filters do not succeed to follow the signals energy envelope and suffer from over or under shoot, ringing, etc.
So, I can understand why an MA filter has an advantage in this case, as well as being phase-linear, but it also shows a latency. Isn't this a problem ?
For instance if I'm interested in designing a compressor, which is based on an EF (envelope follower) for energy level detection. If it would use an MA filter, it could not be used for 'live' compression, just with look-ahead latency compensation. Is this correct?
Or is the FIR latency an issue only for long FIR filters?
I couldn't understand why should someone prefer a MA filter over a simple BiQuad for the case of Delay in the feedback channel.
Wouldn't an MA filter add in-needed latency to this channel?
Is an MA preferred in this case because of phase issues (adding a phased signal to its own original can cause severe frequency cancelations) ?
Any comment is welcomed - thanks !
The MA filter is non-ideal in that its rolloff is poor with low orders. There are some algorithms that might require it where the moving average is important.
I always use simple non-resonant IIR filters for delay feedback.
But you always have the option of making a plugin that uses multiple feedback filter types, testing it, and forming your own conclusions about the sound quality.
There isn't a good comparison for latency because "BiQuad" can be any type of filter (including a 2nd order MA). The delay through the filter for feedforward types always equals the number of delay elements in the algorithm. The "delay" through the IIR filter is another story. Here, you'd have to look at group-delay, the first derivative of the phase response. For high-Q filters, the group delay can be large and can be thought of as related to the ringing-time. But you'd never use a high-Q filter in a sidechain for an envelope follower anyways. There really isn't a perfect comparison. But for a low order MA, it would probably have less latency than a 2nd order, non-resonant LPF's group delay, so it is kind of an "apples and oranges" comparison.
The MA is no more or less recommended for the side-chain of the envelope follower than any other filter, its only a suggestion for use of a basically crappy LPF (for low orders). The audio detector has its own ballistics, so you could have a detector with a 1 second attack time and a 10 second decay time - the filter's latency or group delay won't mean anything in comparison with the detector's ballistics.
Most Users Ever Online: 85
Currently Browsing this Page:
Guest Posters: 1
Newest Members:tomr, Noah, Dave, acv, Vasil992, Vasil92, dowsed, Simple, Chris_1, manyu
Moderators: W Pirkle: 361
Administrators: Tom: 73, JD Young: 80, Will Pirkle: 0, W Pirkle: 361