ProSoundWeb Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: [1] 2 3 ... 5   Go Down

Author Topic: FIR vs IIR  (Read 34616 times)

Ian Stuart

  • Newbie
  • *
  • Offline Offline
  • Posts: 32
FIR vs IIR
« on: April 18, 2011, 08:52:32 am »

Can someone actually explain where FIR and IIR filters are used?
Logged

David Gunness

  • Classic LAB
  • Newbie
  • *
  • Offline Offline
  • Posts: 9
    • Fulcrum Acoustic
Re: FIR vs IIR
« Reply #1 on: April 18, 2011, 11:42:50 am »

IIR filters generally come in small blocks called biquads.  Each biquad can implement one parametric EQ, or one low pass or high pass second order (12 dB per octave) filter.  A fourth order (24 dB per octave) filter requires two biquads.  They use very little processing horsepower and do a good job at replicating the magnitude and phase response of familiar analog filters.  When you change the setting of a single parametric filter, only one biquad has to be changed.  So, they work very well in devices that have a user interface, like a mixing board.

FIR filters usually come in large blocks.  Each large block is capable of implementing a whole bunch of things simultaneously (like several dozen parametric filters, for example), and they are good at certain things that would be extremely difficult with IIR filters, like manipulating the phase response independently of the magnitude response.  To change some small detail of the response of an FIR (like adjust one parametric filter), the entire FIR block has to be recalculated.  Consequently, they tend to be used in applications where their response doesn't have to be changed, and where their unique capabilities are valuable, like loudspeaker processing. 

David Gunness
Logged
David Gunness
Fulcrum Acoustic, LLC

Martyn ferrit Rowe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 334
  • are we there yet?
Re: FIR vs IIR
« Reply #2 on: April 18, 2011, 04:54:35 pm »

Logged

Peter Morris

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1300
Re: FIR vs IIR
« Reply #3 on: April 18, 2011, 10:20:17 pm »

Can someone actually explain where FIR and IIR filters are used?

In very very simple terms, IIR filters are used in the live sound industry to digitally replicate the functions of analogue filters - PEQs and crossovers. 

FIR filters are used to do the same but have two advantages – they can generate filters and crossovers with no relative phase shift and produce very high cross-over slopes.

The disadvantage is the time it takes the FIR filters to do their stuff and it’s not that simple to change their parameters. The time needed is proportional to frequency.  If you need a low frequency crossover with no phase shift it will take a relatively long time. 

The main application for FIR filters in the live sound industry is in speaker processors such as the Lake Contour.  Provided you can tolerate some signal delay (think in terms of 5 - 15 ms) you can produce dramatic improvements in terms of - phase, impulse and magnitude response …. and best of all sound quality, especially if Mr Gunness has been involved ;-)

Peter
« Last Edit: April 19, 2011, 05:55:10 am by Peter Morris »
Logged

Bennett Prescott

  • Full Member
  • ***
  • Offline Offline
  • Posts: 138
  • This text is personal!
    • Bennett Prescott Dot Com
Re: FIR vs IIR
« Reply #4 on: April 18, 2011, 11:49:39 pm »

The main application for FIR filters in the live sound industry is in speaker processors such as the Lake Contour.  Provided you can tolerate some signal delay (think in terms of 5 - 15 ms) you can produce dramatic improvements in terms of - phase, impulse and magnitude response …. and best of all sound quality, especially if Mr Gunness has been involved.

Think in terms of <2ms! If you want to use FIR filtering to do work at low frequencies (say below 800Hz) sure you'll have to pay those kinds of time penalties, but conventional IIR filters are great at those frequencies so use them where they work just fine and don't pay the time penalty. Most of the stuff that FIR can do that IIR can't (or at least not practically) is a higher frequency phenomenon, so you don't need to pay such a steep time penalty.
Logged
-- Bennett Prescott
Director of North American Sales
ADRaudio d.o.o.
Cell: (518) 488-7190

"Give me 6dB and I shall move the world." -Archimedes

Peter Morris

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1300
Re: FIR vs IIR
« Reply #5 on: April 19, 2011, 06:10:31 am »

The main application for FIR filters in the live sound industry is in speaker processors such as the Lake Contour.  Provided you can tolerate some signal delay (think in terms of 5 - 15 ms) you can produce dramatic improvements in terms of - phase, impulse and magnitude response …. and best of all sound quality, especially if Mr Gunness has been involved.

Think in terms of <2ms! If you want to use FIR filtering to do work at low frequencies (say below 800Hz) sure you'll have to pay those kinds of time penalties, but conventional IIR filters are great at those frequencies so use them where they work just fine and don't pay the time penalty. Most of the stuff that FIR can do that IIR can't (or at least not practically) is a higher frequency phenomenon, so you don't need to pay such a steep time penalty

The minimum time that a Lake Contour/ LM26 etc. allows for the implementation of a FIR filter is 1.25 ms in 3 way mode, and 3.15 ms in 4 way mode, add to this a couple of ms for “normal” DSP latency. That gives you a total of between 3 to 5 ms as a minimum.

FWIW 2.5 ms will get you down to about 515Hz @ 48dB/oct. If you allow 12.5 ms you can get down to around 100 hz with a useful cross-over slope

See tables 5.6 to 5.9 for more details 
http://www.dolby.com/uploadedFiles/zz-_Shared_Assets/English_PDFs/Professional/Dolby_Lake_Controller_Manual.pdf

If you can produce a flat phase response and everything else is just right it you will be able to reproduce the original waveform more accurately than ever (see my efforts below with 300Hz square wave) and it will sound very very real especially if you can ge it to  work over most of the vocal range. 

Even if you use IIR filters and just flatten the phase it will cost you time.

I would argue if you are going to delay your FOH to some where on the stage such as the back line or even just to align the top boxes to take account of  path length of a horn loaded sub, why not use the time.

A Gunness Focused KF760 has got a great phase response and needs in the order of 10ms before any sound comes out of the box. Use it with a BH 760 sub (a 10 foot horn) and you are about even. I don’t see the penalty provided it’s a FOH application…. you could just get the audience to stand a few feet closer and it will all come out in the wash.  ;D
« Last Edit: April 19, 2011, 07:27:03 am by Peter Morris »
Logged

Martyn ferrit Rowe

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 334
  • are we there yet?
Re: FIR vs IIR
« Reply #6 on: April 19, 2011, 11:46:43 am »

Hey Guys,
although iir is computationally more efficent at the LF, the distinct advantage of FIR is the ability to "divorce" phase and magnitude. Using fixed-length tap filters gives you a consistent latency.

"move the audience"8)

ferrit
Logged

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 16530
  • Hickory, Mississippi, USA
    • Resotune
Re: FIR vs IIR
« Reply #7 on: April 19, 2011, 12:10:14 pm »

Not to confuse this discussion but divorcing phase response from amplitude response is not always (ever?) a good thing. In terms of speaker management (crossovers and corrective EQ), phase response between drivers very much matters.

It looks like speaker designers who publish presets for loudspeaker crossovers need to also specify IIR or FIR for individual filters used (or actual phase response curves).

JR
Logged
Don't tune your drums half-ass. Listen to what a properly "cleared" drum sounds like.   http://circularscience.com/

Peter Morris

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1300
Re: FIR vs IIR
« Reply #8 on: April 19, 2011, 09:25:28 pm »

Not to confuse this discussion but divorcing phase response from amplitude response is not always (ever?) a good thing. In terms of speaker management (crossovers and corrective EQ), phase response between drivers very much matters.

It looks like speaker designers who publish presets for loudspeaker crossovers need to also specify IIR or FIR for individual filters used (or actual phase response curves).

JR


I think I understand where you are coming from but I’m going to more or less argue the opposite. A perfect system would replicate the input signal exactly. If everything in that signal chain is not minimum phase then the phase and amplitude response will not be directly related (Hilbert transform stuff) and you need to do a few “tricks” to fix things.

Being able to manipulate amplitude and phase separately allows all sorts possibilities. Manipulating the individual coefficients of the delay taps of an FIR filter is the way to go.

Have a look at what these guys are doing … http://www.fouraudio.com/de/hintergrund/inside-the-hd2.html

Once you can get one speaker “perfect” the next trick is to get all the elements of a line array to sum like that across an entire audience. …

The bottom line is what comes out should match what goes in as far as reasonably possible. It must match in amplitude and (relative) time. 

A test that was often seen as the holy grail for a loudspeaker was the ability reproduce a square wave – very few can do that, hence my picture above.  To achieve it I had to manipulate phase independently of amplitude. The resulting frequency, phase and the impulse response have to be relativly perfect.  The only penalty is it takes a bit of time to fix things. In monitor land, it could be a problem, but with FOH by the time I have aligned my top boxes with my horn loaded subs I’m good.

Peter

« Last Edit: April 20, 2011, 03:20:38 am by Peter Morris »
Logged

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 16530
  • Hickory, Mississippi, USA
    • Resotune
Re: FIR vs IIR
« Reply #9 on: April 19, 2011, 11:19:28 pm »

Well no, but I will take the blame for you not knowing where I am coming from.

I am not saying there is no place for FIR "and" IIR, and perhaps even simple delay. My concern is that loudspeaker manufacturer's who publish crossover and corrective EQ presets for their loudspeakers now need to add even more detail namely FIR or IIR to fully characterize what they are saying is their factory presets are.

It's already bad enough with a lack of clear definition for even what Q means in the context of peaking EQ sections, now we need to know even more details about how the filter was implemented.

Don't get me wrong, more control is good, we just all need to be on the same page wrt defining presets. Both amplitude and phase response matters for loudspeaker crossovers.

JR
 
Logged
Don't tune your drums half-ass. Listen to what a properly "cleared" drum sounds like.   http://circularscience.com/

ProSoundWeb Community

Re: FIR vs IIR
« Reply #9 on: April 19, 2011, 11:19:28 pm »


Pages: [1] 2 3 ... 5   Go Up
 



Page created in 0.097 seconds with 22 queries.