Anyone who believes he can get perfect sound using a global FIR filters needs to understand that loudspeakers exhibit power and thermal compression, especially in large setups. Even in small home systems, there' usually enough excursion in the drivers to change their parameters (e.g. BL factor) and consequently the resulting SPL and phase / delay. The room, obstacles, horn flares, cone breakup, diffraction etc. have their own contributions.
Even if speakers are outdoor and there're no room modes, there's still the medium and the related frequency-selective dissipation and propagation distortion of sound waves. This is because sound waves themselves are a result of distortions in air pressure, just like gravitational waves are the result of space-time distortion. There can also be variations in arrival time between various drivers / channels that could render the sound imperfect to the listener.
These are effects that are too detailed to be predicted / modelled / compensated and therefore things no FIR filter can clear.
Even if speakers are outdoor and there're no room modes, there's still the medium and the related frequency-selective dissipation and propagation distortion of sound waves. This is because sound waves themselves are a result of distortions in air pressure, just like gravitational waves are the result of space-time distortion. There can also be variations in arrival time between various drivers / channels that could render the sound imperfect to the listener.
These are effects that are too detailed to be predicted / modelled / compensated and therefore things no FIR filter can clear.
Maybe we shouldn't bother with sound reproduction at all 🙄 .
Sorry, I just can't see how the above post would help analize the sound of a fir corrected speaker vs a non fir corrected speaker.
Sorry, I just can't see how the above post would help analize the sound of a fir corrected speaker vs a non fir corrected speaker.
Indeed but that shouldn't stop sensible people from making good speakers just a little (or maybe a lot depending on the issues) better. Things like making both channels more similar to each other in level and frequency response can have a much greater impact than the filters themselves might imply. These don't have to be FIR or global but it's often easier to do it that way.
Some of them might well be and most of those come down to good system design. Certainly no filter is going to overcome bad design choices. But there are a surprising number of things that can be analysed, measured and corrected for. Like Gunness, at Fulcrum does.
https://www.fulcrum-acoustic.com/wp...dspeaker-transient-response-with-dsp-2005.pdf
Sorry, it wasn't meant to help with analysis of FIR correction, but it mentions the limitations of the method.
Re FIR, there is very much a time and a place for it but a fix all it is certainly not, just another tool in the toolbox, often not the right tool for the job but can do some things extremely well.
A lot of the rest you have mentioned can be mitigated / sidestepped to a degree, some with DSP solutions and others through design such as using drivers through their BL and suspensions linear range, using drivers in their piston range or careful use of damping to smooth the effect of cone breakup, choosing drivers and xover points to minimise thermal compression differences between bands, keeping driver centres within 1/2 a wavelength, reducing polar responce anomolies, preferably less, careful design of waveguides and baffles to reduce reflections and diffraction, the list goes on and on, the more we chip away all goes a long way to revealing the effects of things like time coherence and filter ringing.
Dear fluid & SubSoniks, I was just mentioning that there're a few conditions like linearity, time invariance etc. that a correction filter takes for granted, along with a few errors that linear phase at the acoustic output (as put by john k...) cannot rule out, with absolutely no intentions of offending any FIR worshippers.
If it is your intention not to give offence I would suggest not using phrases like "FIR worshippers".
I was really just trying to take your comments further, not disagreeing 🙂 FIR absolutely needs to be applied carefully with all the above taken into account to not cause more problems than it solves.
I'm trying to read through some of the earlier posts. In the mean time,
https://en.wikipedia.org/wiki/Nonlinear_acoustics
https://en.wikipedia.org/wiki/Nonlinear_acoustics
No it doesn't. It mentions other stuff that is a factor weather one uses FIR or not.
The OP is about using FIR on a per driver base or as a global correction mechanism. When done properly there should be no difference.
But chances are that a per driver solution would be easier to implement and thus easier to get the desired results.
Whatever your intentions were with that post, it had nothing to do with the question on hand. They were just random things that could go wrong in speaker design. Even if one were to correct such a design with FIR filters, it could still reduce the time error of the crossover, regardless of the question if something like that is audible. If other factors create new errors, they would be there without the use of FIR filters too, so what's the point of mentioning them on this thread?
Signed, a FIR worshipper.
P.S. A lot of the time DSP features like this get a bad reputation because it does exactly what it was asked to do by it's user. But it doesn't do what the user wanted it to do. So who's fault is that...
does this mean that a 8th order bessel filter is composed of 4 second order filters with Qs 1.22567, 0.71085, 0.55961, 0.50599 ?
Actually, no, because only a linear system can be compensated /EQed using filters. Non-linearity in audio is insignificant until the amplifier, but the moment you introduce the speaker into the equation (as mentioned earlier), that changes.
That is mostly when they claim "perfect performance" during their promotions / advertisements that possibly raises the expectations of the users beforehand.
But I had no intentions of taking it further, as such errors may not be within our control to compensate them.
Last edited:
So, we're back at the point where it is of no use to reproduce audio 🤔. Thanks for your help.
You should realize that what I said still holds true. Despite the use of non linear speakers. It has been shown in this thread multiple times.
You should realize that what I said still holds true. Despite the use of non linear speakers. It has been shown in this thread multiple times.
Science says wesayso, that only linear time invariant (LTI) systems can be equalised using a filter. Maybe you misunderstand "non-linear". All equalisation assumes linearity and time invariance in the first place.
Last edited:
This is not quite correct. Typically, filters exploiting Volterra series are used to deal with non-linear systems.
And, Dmitrij_S, what percentage of typical loudspeakers use them ? If one starts considering more variables, then the compensation would become unnecessarily complicated, which is not what FIR filters are known for.
Last edited:
The question is right, however, this is another question ! 🙂
P.S.
Short movie by Antonin Novak on filtering/equalizing non-linearities in loudspeaker in real time
Last edited:
Could you point to an verbal / written explanation of the video ?
Meanwhile, I would like to point out a difference here:
* A non-linear compensator expects and models the system to be non-linear in the first place. It therefore (usually) has sufficient knowledge on its non-linear characteristics.
* However, an FIR filter simply expects the loudspeaker to have a fixed magnitude and phase response (like any other filter) at all times, while attempting to correct it.
Meanwhile, I would like to point out a difference here:
* A non-linear compensator expects and models the system to be non-linear in the first place. It therefore (usually) has sufficient knowledge on its non-linear characteristics.
* However, an FIR filter simply expects the loudspeaker to have a fixed magnitude and phase response (like any other filter) at all times, while attempting to correct it.
If I remember correctly, these clips were attached to this conference thesis
https://www.aes.org/e-lib/browse.cfm?elib=20266
The authors use pre-distortion to compensate for nonlinear distortions of periodic signals. Probably, I confused with another article where DSP and Volterra series were used to compensate for the non-linearities of musical signals.
I can agree with this formulation of the filtering problem. However, initial generalazing statement that only LTI system can be equalized using filters, was not entirely correct. It is clear that nonlinear systems require more complex filtering algorithms, nevertheless, such algorithms exist.