I can't understand why you say the discussion dropped into 'FIR general theory', or why you contend real life dsp devices have such actual restrictions.
It's not general theory that minimum phase filters can be embedded in a FIR file.
In fact, probably the greatest single utility of FIR is the ability to embed essentially unlimited IIR EQs into the FIR file.
What analog or even non-FIR dsp doesn't have a comparably anemic max filter capability?
John K already mentioned there is no need for either symmetry or linear phase when using FIR.
I routinely use FIR to implement low latency minimum phase crossovers, for live sound applications. All that's needed is embed a minimum phase crossover and move impulse centering towards the front of the FIR file.
Bottom line...you can put any imaginable filter implementation (that I am aware of) into a FIR file.
That's plain real world...and is routinely done...not theory at all.
I mean, it's just a file of coefficents, right?
Yes, some dsp hardware has tried to bind its customers to their proprietary software for making FIR files.
The god-awful Monkey Forrest software of Four-Audio comes to mind.
But miniDSP for example uses a bin file that many FIR generators can make, and more and more hardware platforms are accepting everyday wav and csv files.
As far as hardware, it just doesn't take much. I've used an old JRiver ID, their little standalone setup on an Intel NUC powered by a little Celeron processor.
That little JRiver NUC could handle 8 channels of convolution, 32K taps per channel. Heck, a raspberry pi is sufficient for the most FIR tasks.
Ime FIR on outputs is easy, no matter what kind of crossovers, individual driver adjustments, etc...that you want to use.
I think Vacuphile said the real constraint...good multi-channel DACs. Amp channels are cheap.
But the sonic benefits of multi-amped speakers properly tuned, greatly over shadow potential DAC differences, and make the FIR on outputs effort well worth it, imo/ime.
Last edited:
Because I have designed several speakers with that real life dsp device; with IIR, IIR+FIR and pure FIR settings. Simple fact which should be easy to understand and accept without need to repeat three times. Please let me know if more repeating is needed. This is only thing in my life so let's copy-paste until we have full 1000.
What he wrongly assumed was that you could do a lot of what he was doing with a passive xover, except for that of the compression, and the adaptive filters resulting from increased output. He also originally thought he was beating Hoffman's Law, when all he was doing was throwing more power at the problem to get more output. You can't beat Hoffman and the laws of physics.
So- in the digital domain, you add +12dB shelving of BSC compensation, for a too-small box example, and you still will clip the amplifiers after the DAC. If you have compressors or adaptive limiters, you're still going to ask more than what the amplifier can effectively give, even if you are helping it limp along by limiting. They just cannot sustain that amount of boost without having enormous power reserves available. 300W is not enough.
For my previous example...
As to the thought my Parallel LCR was not at the right frequency range as compared to the Series LCR across it, this thought is incorrect. I am no amateur here. The notches were correctly targeting the right frequency. Only when I prevented the resonant frequency or removed it from the signal by allowing it a different path to ground was the problem removed. This is referencing an energy storage spike in the impedance at 1.5kHz or so, and a magnitude of 4-5 ohms above nominal. This is not a reference to the Fs of the driver itself, or breakup from an average metal cone. The cone is flat and circular, and therefore will have an issue where the diameter of the cone frequency occurs.
So- if you defeat this in the digital realm, you're still minimizing the amount of signal produced at this frequency in particular. Does this mean another frequency won't excite it by a harmonic? I still don't see how this will ultimately stifle the problem.
While I know that approaching zero impedance magnitude is likely not theoretically possible with all the wire and coil resistances in said LCR, it still has 'negative energy' in that direction to do as I've stated. The lowpass then swamps the approach of zero due to the inductive rise. Even if a harmonic then excites this blemish, it still is remedied by said notch.
It takes 3 parts to fix the issue, when the multiple amplfiers, DSP, and DACs are all required to do it the other way. It costs a lot more money for all of that gear to do it that way.
I've had these discussions before about the active/passive preferences, so that is effectively an opinion of the builder, but it does not make one method better than the other.
Back to andy2's intended agenda...
Wolf
MOTU.com - Overview
This is what I plan on using. Some have had good results with similar motu products. $600.00 for 8-10 outputs.
My only "claim" was that you can turn any old ugly pig into a straight line FR with linear phase (The "Lipstick") by DSP. If it will sound good is an other matter.
//
Good to know, thx. Looks like a nice unit.
Another decent $600 solution could be the Behringer XR-18 mixer..if a person wanted a more multipurpose box.
I used a x-32 for a while as an interface to JRiver convolution and it worked great. Outputs measured good, sounded good. XR-18 is supposed to have same processing. The PC/tablet/phone control apps for the Behringer mixers are outstanding.
One example is Genelec S360A
Phase linearisation is in the input, before XO and EQs which are probably IIR.
Phase linearisation is in the input, before XO and EQs which are probably IIR.
Interesting link to AudioXpress, had missed this article.
The revelation of this truth, however, will be a sore disappointment to some others. It may not even change behaviour Sunk cost fallacy | behavioraleconomics.com | The BE Hub
The revelation of this truth, however, will be a sore disappointment to some others. It may not even change behaviour Sunk cost fallacy | behavioraleconomics.com | The BE Hub
One more theoretical study about phase linearization with input FIR. This is 3-way with LR8 IIR crossover at 250 and 2500 Hz. Phase responses matched with two 1st order IIR all-pass filters.
Maximum phase mismatch at -60 dB below nominal sensitivity is about 5 degrees while it was about 20 degrees with IIR LR4 version shown earlier. Both good enough by far for the purpose.
Maximum phase mismatch at -60 dB below nominal sensitivity is about 5 degrees while it was about 20 degrees with IIR LR4 version shown earlier. Both good enough by far for the purpose.
It does, with OpenDRC-DI upfront. The downside is it needs another hw step after it, i.e. the actual crossover.
Your sunk-cost-fallacy link made me laugh, thanks !
Because that's the exact assessment I thought when I read the article a week or so ago
in Voice Coil I think. It read to me like Vance has such a long term investment in his beliefs that he looked for, and cited, evidence to support them.
I get disgusted with myself when I look at all the audio beliefs I've held onto over my life...the purist stage (ie no eq, no tone control, least is best, class A only, etc)...the analog signal stage (digital just can't make true analog waveforms lol)....the maybe digital is OK for recording but not for processing or crossovers stage, more lol)...the phase matters stage...the phase doesn't matter stage ? haha)
Yep. You can get carried away with the technology. At some point you realize that you aren't listening to the music. Today nothing pleases me more than to go out for a jog and listen a bunch of mp3 audio tracks stripped from YouTube videos played on my SanDisk Clip. But wait, those ear buds don't have a crossover. TP? LOL.
None of this stuff makes speakers better, just different. And that's even more to the point. There is a difference between "I heard a difference", and "it sounds better".
I like your style - short and to the point. At first though I thought it was like equivalent of giving the middle finger
Maybe that's why you're an engineer It's true that with active DSP, you can realize a perfect 0 phase shift, proper step response and all. And you can also have LR2, LR4, and so on. Now you start having all sort of different implementation and all of which can produce a proper step response, 0 phase shift and all. It seems like if you use miniDSP equipment, there is a Python program called "RePhase" that can do all the works (or most of it), but I think the constraints is with "RePhase" you can only use LR type filters (LR2, LR4 …).
So the question to ask is "Are all time-coherent, 0 phase shift designs are created equal?" I mean they all will measure more or less the same - 0 phase shift, proper step response, so why would one is the better of the others? I understand the question of "Is lower order filter better than higher order filter?" has been beaten to death but here you go again.
To think that they all sound the same, or they don't make any difference is pretty boring. I doubt God would create such a boring world. Ain't no utopia for sure. In that respect, I think HE came from the marketing department lols and probably the most devilish among the siblings.