In one of the articles I linked above to better illustrate the lobing problems which simply disappears with a brick-wall FIR filter:
This is his set up with two full range drivers spaced 1 wavelength apart at crossover:
And this is the polar response comparison of a normal DSP LR24 slope (still very steep) versus a 'brick-wall' FIR filter:
Figure 14 – The polar response at 1 kHz (1/3-oct, 5-degree angular resolution) using the LR24 crossover (left) and the linear phase brickwall crossover (right).
Whilst I am still undecided ultimately on the absolute audibility of phase and therefore the merits of linear phase versus quasi-transient-perfect crossovers etc, I can see no down side to the removal of the lobing issues in so many crossover designs.
I thick it is this aspect of FIR filtering which is often overlooked.
This is his set up with two full range drivers spaced 1 wavelength apart at crossover:
And this is the polar response comparison of a normal DSP LR24 slope (still very steep) versus a 'brick-wall' FIR filter:
Figure 14 – The polar response at 1 kHz (1/3-oct, 5-degree angular resolution) using the LR24 crossover (left) and the linear phase brickwall crossover (right).
Whilst I am still undecided ultimately on the absolute audibility of phase and therefore the merits of linear phase versus quasi-transient-perfect crossovers etc, I can see no down side to the removal of the lobing issues in so many crossover designs.
I thick it is this aspect of FIR filtering which is often overlooked.
An FFT has a length, though. That length means that the bandwidth of each "bin" can be no narrower than the SampleRate divided by the FFTlength, and you won't have any filter edges happening steeper than the edge of that bin is.
Last edited:
Only answer if you know, don't speculate, the man asked an exact question - you are waffling and have no point nor anything to substantiate what you are saying.
Are we not getting hung up here on the concept of an 'absolute brick wall filter'.
Surely this is one for all intents and purposes (and apparently with no audible pre-ringing and flat phase......):
The frequency response magnitude of a 96 dB/oct crossover network. These filters sum to produce a perfectly flat magnitude response.
The Group delay response
This is 'only' a 96dB/Oct XO, as I linked in the 2008 paper above - apparently as long as it is kept 'below about 600dB/Oct' the pre-ringing is inaudible.
And who cares about 250 msec of latency when you are at home listening to your favorite track??
Surely this is one for all intents and purposes (and apparently with no audible pre-ringing and flat phase......):
The frequency response magnitude of a 96 dB/oct crossover network. These filters sum to produce a perfectly flat magnitude response.
The Group delay response
This is 'only' a 96dB/Oct XO, as I linked in the 2008 paper above - apparently as long as it is kept 'below about 600dB/Oct' the pre-ringing is inaudible.
And who cares about 250 msec of latency when you are at home listening to your favorite track??
Last edited:
I don't know but Neyquist proved that you need two sample per cycle and a bit so 44.1kHz springs to mind. Processors are much faster these days and to make a brick-wall filter at any audio frequency will be possible without breaking a sweat.
This is a good answer in my opinion, it gives some negatives but also positives which has a logical outcome. Worth trying it. These are switched capacitor eight order elliptical filters available from MAXWELL which I have discussed here some years before and the make absolutely great cross-over filters. I don't experience any of the speculations that was posted here.
Sure but I do know, so I know how deep the rabbit hole goes. Do you think that it is possible to answer the OP's question succinctly? A simple question of "why" can sometimes require a long and complicated answer that draws understanding from multiple fields on a non-trivial level. Even aspects of what he is asking would need quite a long an involved answer to properly present and frame the topic. That was what I was trying to convey in my answer, not some kind of "I know better than you and I'm not going to tell you" sort of response.
Before making your next post you might glance down at your very own signature and see how it measures up to the 'regards' embodied in the words that you wrote.
I'm glad to hear that you asked this question. Here are some suggestions on reading material on these kinds of topics. I hope that others will add their own suggestions:
Mathematics - you will need to understand complex numbers well and other mathematical stuff related to audio.
Complex Numbers: https://en.wikipedia.org/wiki/Complex_number for a start, then Google
Fourier Transform (continuous and discrete) - https://en.wikipedia.org/wiki/Fourier_transform then Google
Hilbert Transform (relates amplitude to phase response) - The law of Bode and the Hilbert transformation and next ref:
"Kramers-Kronig, Bode, and the Meaning of Zero" great paper about the HT by Bechhoefer, currently available on line at http://www.sfu.ca/chaos/papers/2011/KK-Bode-MeaningZero.pdf
Filtering - I learned about filtering following the progression: analog active filters -> IIR digital filters --> FIR filters. IMO this approach lets you ease in the subject. I bought some books on Ebay, and then from there knew what to look for on the web. Some references:
Stephenson "RC Active Filter Desigin Handbook" (excellent)
Valkenburg - "Analog Filter Design"
NOTE: there are many, many good books about analog active filter design
online: "The Scientist and Engineer's Guide to DSP" at The Scientist and Engineer's Guide to Digital Signal Processing
DSP Guru web site - dspGuru.com | Digital Signal Processing Central
Comparisson IIR and FIR filters | Crazy Audio (lightweight but informative)
Acoustics - some reference I turn to:
Beranek - Acoustics
Kinsler - Fundamentals of Acoustics
Toole - Sound Reproduction (more down to earth, info on what "sounds good" in a loudspeaker, etc.)
Linkwitz "Conversations with Fitz" (an ongoing free-form lecture on loudspeaker design, acoustics, hearing, etc.) at Linkwitz Lab - Loudspeaker Design
This is a start, and others can chime in with more.
The Scientist & Engineer's Guide to Digital Signal Processing, 1999 | Education | Analog Devices
I already linked a chapter, but it's a nice resource that gets more to the nitty gritty sooner.
I already linked a chapter, but it's a nice resource that gets more to the nitty gritty sooner.
Isn't this apples and oranges? Switched cap filters are minimum phase analogs of ordinary R/C filters. FFT filters can be both linear and minimum phase and you can also sometimes dramatically reduce pre-ringing by introducing small fractional sample delays.
As well as the lung fish the Arapaima that lives in the Amazon breathes air. This used to give it a very good hunting advantage as it could hang around in fairly stagnant water. Sadly man realised that 80kg of tasty would come to the surface every 20 minutes where they could hit it and catch it so now they are in decline.
Because sound doesn't come in discrete frequencies, or with meta-data attached to it to identify a "frequency." Instead all we have to deal with is change in amplitude over time.
So in your world, what happens to the signal at 2,000.005 Hz? What about 2,000.1 Hz?
At best, digital filters are like catamaran, bobbing in the ocean. The wider the distance between the hulls, the lower the frequency they respond to, and the narrower the hull distance, the faster, but at no point is the contribution of "out of band" signals zero. There will always be some bobbing of the hull caused by every wave that passes underneath.
Another problem, is that a crossover's influence (passive or active) ADDS to the driver's response, it doesn't substitute for the driver's response.
Perhaps your problems are best solved by line sources or "full-range" drivers. 😀 They do away with the question of crossover points entirely and leave what remains to be solved in the domain of DSP. Phase, and amplitude compensation.
The design of any quality loudspeaker requires the simultaneous solution of many dimensions. Price, aesthetics, efficiency, dispersion, distortion, on and on and on. This is why there are so many speaker designs out there. The crossover may only be a small part of the overall problem set. 🙂
Best,
Erik
So in your world, what happens to the signal at 2,000.005 Hz? What about 2,000.1 Hz?
At best, digital filters are like catamaran, bobbing in the ocean. The wider the distance between the hulls, the lower the frequency they respond to, and the narrower the hull distance, the faster, but at no point is the contribution of "out of band" signals zero. There will always be some bobbing of the hull caused by every wave that passes underneath.
Another problem, is that a crossover's influence (passive or active) ADDS to the driver's response, it doesn't substitute for the driver's response.
Perhaps your problems are best solved by line sources or "full-range" drivers. 😀 They do away with the question of crossover points entirely and leave what remains to be solved in the domain of DSP. Phase, and amplitude compensation.
The design of any quality loudspeaker requires the simultaneous solution of many dimensions. Price, aesthetics, efficiency, dispersion, distortion, on and on and on. This is why there are so many speaker designs out there. The crossover may only be a small part of the overall problem set. 🙂
Best,
Erik
Last edited:
posts deleted, infractions given, for the nth time, attack the idea not the poster....
Haha, another audio war starts at diyaudio. 🙂
The simple issue with brick wall filters is that infinite steepness equates to infinite group delay or ringing. That's the maths.
Every speaker is compromised. More of this, less of that. Maybe we can do some tricks with computers, but it's all a trade off.
Nature just knows when you split the signal path in two
But some solutions work well enough. Study "Classic Monitor Designs" if interested enough.
The simple issue with brick wall filters is that infinite steepness equates to infinite group delay or ringing. That's the maths.
Every speaker is compromised. More of this, less of that. Maybe we can do some tricks with computers, but it's all a trade off.
Nature just knows when you split the signal path in two
But some solutions work well enough. Study "Classic Monitor Designs" if interested enough.
Steve, is that "Classic Monitor Designs" an actual book or paper title? Did a search and didn't find anything.
Thanks
Thanks
"Classic Monitor Designs" is a diyaudio.com thread that went way over the heads of a lot of people here, especially including the forum hyenas, and we know who those characters are.
Google it. I am on mobile so can't give you the link. The idea is that there is an awfully good repertoire of good designs that satisfy. Nothing is perfect of course. We just seek "good enough" IMO.
Really, people who think there is some perfect speaker just don't understand how it works. Now maybe perfect headphones are doable.
Google it. I am on mobile so can't give you the link. The idea is that there is an awfully good repertoire of good designs that satisfy. Nothing is perfect of course. We just seek "good enough" IMO.
Really, people who think there is some perfect speaker just don't understand how it works. Now maybe perfect headphones are doable.
Got it! Thanks !
That begs a question but I just don't know how to frame it in few words. Cabinet, crossover and room acoustics out of the way, does the design of the headphone get credit for not having to deal with them? Is that the neighborhood ?
Thanks
That begs a question but I just don't know how to frame it in few words. Cabinet, crossover and room acoustics out of the way, does the design of the headphone get credit for not having to deal with them? Is that the neighborhood ?
Thanks
- Status
- Not open for further replies.