Which FIR processors suitable for install loudspeaker management will allow downsampling by channel?
I'm researching platforms with FIR processing for phase correction. Many comments are made about the available number of taps for lower frequencies and how there never seems to be enough. But this can be done with significantly fewer taps by downsampling the low frequency channels in the processor, correct? Looking at some different platforms however (hoping for "affordable" ones like Xilica Solero) I don't see any reference to this. Not sure if it's unavailable or just weak documentation.
I'm researching platforms with FIR processing for phase correction. Many comments are made about the available number of taps for lower frequencies and how there never seems to be enough. But this can be done with significantly fewer taps by downsampling the low frequency channels in the processor, correct? Looking at some different platforms however (hoping for "affordable" ones like Xilica Solero) I don't see any reference to this. Not sure if it's unavailable or just weak documentation.
as a practical solution could you use IIR filters for the low frequency?
Mixed rate processing is possible and widely used in digital radio but I am unaware of any processor that explicitly exposes the ability to change sample rate. It may be that proprietary 'brick wall' crossover filter types use multirate processing.
Kind of unrelated but I wonder what this is if it isn't an FIR filter:
http://linea-research.co.uk/wp-cont...ssets/Tech Docs/LIR_LinearPhaseCrossovers.pdf
(IIR with phase correcting all pass FIR?)
Mixed rate processing is possible and widely used in digital radio but I am unaware of any processor that explicitly exposes the ability to change sample rate. It may be that proprietary 'brick wall' crossover filter types use multirate processing.
Kind of unrelated but I wonder what this is if it isn't an FIR filter:
http://linea-research.co.uk/wp-cont...ssets/Tech Docs/LIR_LinearPhaseCrossovers.pdf
(IIR with phase correcting all pass FIR?)
FourAudio has one that might do it. http://www.fouraudio.com/data/HD2_Manual.pdf
Can't tell for sure from the manual, but the have technical papers touting down-samplimp for low freq work.
At any rate I'm sure it's plain awful expensive.
But I think in PA we are still forced to use IIR down low, as downsampling doesn't change latency. Just changes processing power needed.
The Linea LIR filters are essentially linear phase LR24's. They have a really slick implementation, in that you can dial in the xover freq in real time, and the processor automatically keeps time alignment between all output channels linked together.
LIR xover has about 12-15ms latency at 100Hz, and heading to much lower latency as freq increases. Just like roll your own LR 24 FIR tap requirements.
Can't tell for sure from the manual, but the have technical papers touting down-samplimp for low freq work.
At any rate I'm sure it's plain awful expensive.
But I think in PA we are still forced to use IIR down low, as downsampling doesn't change latency. Just changes processing power needed.
The Linea LIR filters are essentially linear phase LR24's. They have a really slick implementation, in that you can dial in the xover freq in real time, and the processor automatically keeps time alignment between all output channels linked together.
LIR xover has about 12-15ms latency at 100Hz, and heading to much lower latency as freq increases. Just like roll your own LR 24 FIR tap requirements.
Kind of unrelated but I wonder what this is if it isn't an FIR filter:
http://linea-research.co.uk/wp-content/uploads/LR%20Download%20Assets/Tech%20Docs/LIR_LinearPhaseCrossovers.pdf
(IIR with phase correcting all pass FIR?)
They are being disengenous with their latency graph - you can make FIR filter tap length optimal to each possible frequency and get pretty much the same smooth curve they do. My suspicion is they use a tunable sample rate and a single optimized filter, either digital or analog FIR (switched capacitor array or something like that).
Last edited: