hi,
things are starting where they end...
considering loudspeakers ARE analog devices we have to play with analog rules.
grasso says and I do agree!
"In the analogue world resonance is always connected to time delay, the impulse response is not symmetrical. But linear-phase nonlinear-amplitude filters have a symmetrical impulse response, they do "pre-ringing". A crossover built of linear-phase filters is not detectable only in case that the sum of lowpass and highpass is 1, unity gain. Linear phase needs linear amplitude. Interference due to the travel time of sound in air (340m/s), in particular the time to travel the distance between several loudspeakers which transmit the same frequency, messes up everything. Try to archieve an even amplitude response at all radiation angles of the speakers! Do not overstrech it! Avoid preringing! Analogue is akin, digital is dangerous! An anecdote is that nowadays loudness, that is average energy divided by peak amplitude, is considered a quality criterium, so the sound of Pop/Rock CDs and broadcasts is coloured by compressors. And many engineers think that they could improve the sound of multiband compressors, in explanation end-of-the-chain loudness maximizers, and exiters, say overtone generators, by replacing minimum-phase by linear-phase filters. Oh my!"
One other thing, FIR filter processing is not the matter of cpu power, it can be easily implemented on elder MMX or 3Dnow! processors.
Using an AMD K6-2 500MHz, 6 simultaneously running FIR filters (sample fs=44100Hz), each wrapping 127 coeffs, using 6% CPU power!
now, my opinion is, high quality crossover frequencies ARE analog devices, due to the fact of fundamental law of nature.
regards, millim
things are starting where they end...
considering loudspeakers ARE analog devices we have to play with analog rules.
grasso says and I do agree!
"In the analogue world resonance is always connected to time delay, the impulse response is not symmetrical. But linear-phase nonlinear-amplitude filters have a symmetrical impulse response, they do "pre-ringing". A crossover built of linear-phase filters is not detectable only in case that the sum of lowpass and highpass is 1, unity gain. Linear phase needs linear amplitude. Interference due to the travel time of sound in air (340m/s), in particular the time to travel the distance between several loudspeakers which transmit the same frequency, messes up everything. Try to archieve an even amplitude response at all radiation angles of the speakers! Do not overstrech it! Avoid preringing! Analogue is akin, digital is dangerous! An anecdote is that nowadays loudness, that is average energy divided by peak amplitude, is considered a quality criterium, so the sound of Pop/Rock CDs and broadcasts is coloured by compressors. And many engineers think that they could improve the sound of multiband compressors, in explanation end-of-the-chain loudness maximizers, and exiters, say overtone generators, by replacing minimum-phase by linear-phase filters. Oh my!"
One other thing, FIR filter processing is not the matter of cpu power, it can be easily implemented on elder MMX or 3Dnow! processors.
Using an AMD K6-2 500MHz, 6 simultaneously running FIR filters (sample fs=44100Hz), each wrapping 127 coeffs, using 6% CPU power!
now, my opinion is, high quality crossover frequencies ARE analog devices, due to the fact of fundamental law of nature.
regards, millim
This is an extremely simplistic view of the situation. ANY system that has multiple independent paths (ie multiple drivers) for the same signal will experience off-axis response anomalies - this is normally just called 'lobing' for conventional analog/IIR filters, but that doesn't really capture the full effect. It is true that the symmetric nature of purely linear phase filters can induce pre-ringing that will only cancel on-axis (unless you're using co-axial drivers), I think it is extremely naive to automatically assume that this invalidates the approach.
In truth, it is likely that neither purely linear-phase nor purely minimum-phase topologies are the optimal approach for a digital system, but that something like a subtractive-delay or hybrid topology will prove to be more satisfying, albeit at the expense of more complexity to generate the filters.
Note that this addresses xover filters only. For room correction a mixture of minimum and linear phase is pretty much unavoidable due to the varying natures of the room effects.
BTW - 127 tap FIR filters are awfully short. I find it unlikely that you'll get 'high end' results out of such short filters.
In truth, it is likely that neither purely linear-phase nor purely minimum-phase topologies are the optimal approach for a digital system, but that something like a subtractive-delay or hybrid topology will prove to be more satisfying, albeit at the expense of more complexity to generate the filters.
Note that this addresses xover filters only. For room correction a mixture of minimum and linear phase is pretty much unavoidable due to the varying natures of the room effects.
BTW - 127 tap FIR filters are awfully short. I find it unlikely that you'll get 'high end' results out of such short filters.
- Status
- Not open for further replies.