rePhase, a loudspeaker phase linearization, EQ and FIR filtering tool

It is easy to limit the Numbers of Filters to 17, so that they will fit to a single Bank(Screenshot) and I never found Notch Filters in automated Corrections.
Usually I do not get more then 5 -6 Filters..

REW Filters Sets are more a "Suggestion" to me and in most of the Times not final, but they are a very good start and make Work faster....

Regards
 

Attachments

  • Screenshot - 25.04.2016 - 08:18:15.png
    Screenshot - 25.04.2016 - 08:18:15.png
    40.1 KB · Views: 605
Last edited:
It is easy to limit the Numbers of Filters to 17, so that they will fit to a single Bank(Screenshot) and I never found Notch Filters in automated Corrections.
Usually I do not get more then 5 -6 Filters..

REW Filters Sets are more a "Suggestion" to me and in most of the Times not final, but they are a very good start and make Work faster....

Regards
I will start implementing an import functionality for the next release.
 
Hi guys, total newb on this interesting concept here, I have mainly one question to this:

I'm into club-sound systems and currently run a 3-way system, I've been reading up a bit on miniDSP site and forum and it seem that the best way to get a Phase-linear three way system with the rePhase is to use for example a DCX-2496 with IIR filters, and use the filter linearization in the rePhase plugin to linearize the total stereo signal with a openDRC box before the DCX-2496, using a measurement microphone in front of the sound system?

What made me a bit confused is the title of this thread, "fir filtering tool" which made me think I could implement a three way FIR filter with balanced outputs to be able to ditch the IIR processor totally, but the multi-channel openDRC boxes seem to only supply unbalanced connectors. (It seem that Pos also adressed this "problem" in a post about "blurring the original purpose of the software" on the miniDSP forum) which I guess is what I experienced :)

So is this for example the correct signal chain to use it?: mixer - openDRC - dcx2496 - amp - speaker - measurement mic

And then manually correct the phase in the program with the filter linearization tool?

Another question: would this method yield equal or similar results as implementing FIR filters to begin with? With for example the Lake LM26.

Thank you for everything you provide to the DIY community!
 
Last edited:
Hello Osse

rephase will let you generate FIR that can correct an existing crossover, or implement the crossover. As you mentioned if you want hardware solutions there is not currently that much of a choice on the market.
openDRC-D8 and 2x4HD are good solutions if the number of taps they offer is enough for your needs and you can accommodate unbalanced outputs.
Another solution is to use several openDRC (either DA or DI with a DAC, which is the solution I use in my system) and split the input signal between them.
Najda in another good integrated solution, but with unbalanced outputs again.
Other FIR devices will probably come to market in the near future, like for example Tranquility Bass's solution, which looks pretty impressive.

There are other FIR units, especially on the pro market, like the EV DX46, some Xilica units, BSS and Crown units, and Dolby Lake as you mention. The problem with these units is that FIR capabilities are limited to textbook crossovers (brickwall shaped by windowing, or linear phase LR types), you cannot import externally generated FIR (in a documented format at least, or maybe I missed something).
Textbook linear phase filters are of no use in a real world situation (unless you resort to sharp brickwall filters, which bring their own problems) because you have to take into account the existing rolloff (and assiociated phase shift) of your drivers. This *can* be done using IIR, but once again these units do not offer the right tools for that (linkwitz transform/asymetric shelving) and there is only so much you can do with PEQ alone.
The Ev DX46 is particularly frustrating in this regard, as it is a fixed point unit and the IIR section takes place *before* the FIR one, making it impossible to use large positive EQs without clipping the whole thing...
 
Last edited:
Thanks for your response Pos.

What is intriguing to me about the openDRC concept is the ready-to-use box and balanced connectors(the DI,AN,DA ones) and not only a PCB-board like with for example the Tranquility bass solution.

It doesn't matter to me if i generate FIR filters or compensate IIR filters with a FIR phase compensation tool as long as it gets the job done and is cheaper than the LM26(the openDRC boxes are very humbly priced).

I've only played around with IIR filters in my LMS and can get my system to sound really good with it, but I'm curious what a FIR linear phase and EQ would bring to the table as icing on the cake.
 
Even for minimum-phase corrections (such as EQs), and if you have enough taps at hand to do the job, FIR has a theoretical advantage over IIR: quantization noise can rise quickly with IIR, due to the recursive nature of its biquad implementation, whereas FIR convolution implementations (especially direct temporal domain one, like in the openDRC and most other hardware units) are much more simple and will not accumulate errors.

And then of course you have the linear-phase thing :)
 
  • Like
Reactions: 1 user
Hi guys, total newb on this interesting concept here, I have mainly one question to this:

I'm into club-sound systems and currently run a 3-way system, I've been reading up a bit on miniDSP site and forum and it seem that the best way to get a Phase-linear three way system with the rePhase is to use for example a DCX-2496 with IIR filters, and use the filter linearization in the rePhase plugin to linearize the total stereo signal with a openDRC box before the DCX-2496, using a measurement microphone in front of the sound system?

What made me a bit confused is the title of this thread, "fir filtering tool" which made me think I could implement a three way FIR filter with balanced outputs to be able to ditch the IIR processor totally, but the multi-channel openDRC boxes seem to only supply unbalanced connectors. (It seem that Pos also adressed this "problem" in a post about "blurring the original purpose of the software" on the miniDSP forum) which I guess is what I experienced :)

So is this for example the correct signal chain to use it?: mixer - openDRC - dcx2496 - amp - speaker - measurement mic

And then manually correct the phase in the program with the filter linearization tool?

Another question: would this method yield equal or similar results as implementing FIR filters to begin with? With for example the Lake LM26.

Thank you for everything you provide to the DIY community!

Hi Osse, near total noob here too, other than I've been playing with rePhase and some mini-dsp units for a month or two..
I'm trying to do the same thing you are with some 4-way PA gear (3-way main and sub).

I'll post comments with the hope they will be supported or corrected as need be... IOW, don't bank on anything I say, haha

As you allude, it seems there are two main paths to phase flattening.
One is too use our IIR equip as usual, and insert miniDSP / rephase (may i term this 'MDR' ?) into the chain as a global phase correction.
The second is to use MDR as the complete crossover solution on its own, and perform driver by driver phase flattening.

From what I understand, either will work nearly identical for a single point in space.
But, using MDR as a complete driver by driver solution, is superior for multiple points in space.

Reason...well, when thinking about on and off axis response like we have to with club/live sound gear...
We know that we have ultimately to measure off axis response at different x-over frequencies and typologies within the possible range of crossover freqs, to find optimal off axis response. On axis is relatively easy.
If we use MDR as a global correction, trying different x-over freqs and typologies without rerunning an entire new global response for each attempt is invalid, because we flattened phase based on the combined response of drivers at a previously given freq and topology..
If we use MDR driver by driver, we have a lot of room to move x-over freq and topology around as we search for optimal, because we should have good summation.....because each driver is flat with the range we deem possible...

I'm going through the MDR driver by driver route right now.
I hear you regarding the lack of balanced outs on the minDSP units.
Second bummer, even one of the OpenDRC which does have a pair of balanced outs, isn't capable of pro voltage.
I'm solving this temporarily, by running the miniDSP outs into an x-32 rack on the way to the amps. I hate to tie up the rack for this duty, but right now the combo of MDR and the rack seem to be a poor man's LM26 (that's a hell of alot more powerful/capable than a LM26 haha)
 
From what I understand, either will work nearly identical for a single point in space.
But, using MDR as a complete driver by driver solution, is superior for multiple points in space.

If your original IIR crossovers are complementary (eg acoustical symmetrical LR) then linearizing the phase will bring the exact same result as if you did build the crossover using acoustical LR slopes with linear phase, on and off axis.

The advantage of going "full-blown" FIR is that you can choose different complementary acoustical filter slopes than the one you could achieve with IIR (overlapping, very steep, asymmetrical, H-K, etc.). It will also be much easier to design, as you noted: phase tracking becomes a lot easier when the only thing you have to shoot for is a flat phase trace, and you also don't have to worry anymore about the effect of one crossover point phase shift on the others.

... and IMHO when applied with a direct convolution FIR is conceptually (and effectively) so much cleaner than a recursive IIR implementation :D
 
Last edited:
  • Like
Reactions: 1 user
If your original IIR crossovers are complementary (eg acoustical symmetrical LR) then linearizing the phase will bring the exact same result as if you did build the crossover using acoustical LR slopes with linear phase, on and off axis.

The advantage of going "full-blown" FIR is that you can choose different complementary acoustical filter slopes than the one you could achieve with IIR (overlapping, very steep, asymmetrical, H-K, etc.). It will also be much easier to design, as you noted: phase tracking becomes a lot easier when the only thing you have to shoot for is a flat phase trace, and you also don't have to worry anymore about the effect of one crossover point phase shift on the others.

... and IMHO when applied with a direct convolution FIR is conceptually (and effectively) so much cleaner than a recursive IIR implementation :D

Yes, thank you POS

Your posts, as much as your program, have helped me see this.
 
If your original IIR crossovers are complementary (eg acoustical symmetrical LR) then linearizing the phase will bring the exact same result as if you did build the crossover using acoustical LR slopes with linear phase, on and off axis.

The advantage of going "full-blown" FIR is that you can choose different complementary acoustical filter slopes than the one you could achieve with IIR (overlapping, very steep, asymmetrical, H-K, etc.). It will also be much easier to design, as you noted: phase tracking becomes a lot easier when the only thing you have to shoot for is a flat phase trace, and you also don't have to worry anymore about the effect of one crossover point phase shift on the others.

... and IMHO when applied with a direct convolution FIR is conceptually (and effectively) so much cleaner than a recursive IIR implementation :D

Hi POS, ..like I said, I am seeing the equivalence you speak of in the two methods ....for electrical filters.

What I'm still wrestling with is, can equivalence exist for correcting the drivers' acoustical phase response too?
IOW, I know each driver has an inherent acoustical phase curve. Can their summed phase responses be flattened, or must each driver be flattened individually?

I guess the real question is, does acoustical phase summing even exist?
Or is it that waves stay separate in phase, but sum in magnitude based on their varying separation?

I can see there is magnitude summing depending on phase angle, I just can see what phase summing means....or how you could flatten 'what I'm wondering if even exists' :spin:
 
Hi Mark

In a properly implemented acoustical LR crossover both acoustical filters are perfectly complementary and sum to a flat magnitude response (on axis at least).
When temporally aligned each driver encounters the same phase shift, which is also reflected as an allpass in the final summed response. This is called phase tracking and means the two drivers are in phase with each other.
If you then "linearize" the phase of that summed response with an inverse allpass you get a perfect linear phase crossover, exactly identical to what you would get with two linear phase "LR" acoustical filters.

But getting a perfect phase tracking is not that easy in practice when you have several crossover points, because of the influence of other crossover points. In this situation you have to "replicate" lower filters as if the crossovers were chained (cf Linkwitz link I posted a few posts above).
 
Hi again, thank you for your response mark100! Have you tried to globally correct the phase incoherency with your soundsystem aswell? Would be awesome to hear your reports and findings if it benifits your system.

Right now I'm using Keystone tapped horns which are really good, but they have a bit of phase shift and I wonder how they would sound corrected and with IIR filters phase corrected out aswell.
 
Last edited:
Hi Mark

In a properly implemented acoustical LR crossover both acoustical filters are perfectly complementary and sum to a flat magnitude response (on axis at least).
When temporally aligned each driver encounters the same phase shift, which is also reflected as an allpass in the final summed response. This is called phase tracking and means the two drivers are in phase with each other.
If you then "linearize" the phase of that summed response with an inverse allpass you get a perfect linear phase crossover, exactly identical to what you would get with two linear phase "LR" acoustical filters.

But getting a perfect phase tracking is not that easy in practice when you have several crossover points, because of the influence of other crossover points. In this situation you have to "replicate" lower filters as if the crossovers were chained (cf Linkwitz link I posted a few posts above).

Hi Pos,
Thx. Read the Linkwitz link and understand the influence of other crossover points...like you say, all the more reason for linear-phase crossover filters. :)
I'm pretty familiar with phase tracking...been using smaart to align phase traces, subs to main, for a while.
All in all, electrical crossovers are making sense....no real difficulty there..

I'm still not quite so clear with the mechanical part....ie the drivers' raw acoustic response....and corrections for these responses.

I have a hard time seeing how a global correction can correct the phase of each driver individually, through the x-over region. I can see how global can correct the summed response easy enough..
But it still seems like a summed correction is suboptimal to having each of the drivers flattened first..

Sorry for being so dense....
But hey, I really do think I get compensate now :D
 
Hi again, thank you for your response mark100! Have you tried to globally correct the phase incoherency with your soundsystem aswell? Would be awesome to hear your reports and findings if it benifits your system.

Right now I'm using Keystone tapped horns which are really good, but they have a bit of phase shift and I wonder how they would sound corrected and with IIR filters phase corrected out aswell.

Hi Osse, no I haven't tried global correction yet....intend to though...still sorting out understanding how it works as you can tell :confused:

I'm using Labhorns and JTR OS subs, also horn loaded. Correcting the IIR filters is easy enough, although correcting a HP takes a lot of taps and latency. I haven't tried to correct the subs acoustic response yet.
All my effort so far has been directed at a DIY main...here's a post i made yesterday in another forum re progress on it...https://soundforums.net/threads/12075-60-Degree-DIY-Mid-Hi?p=100676&viewfull=1#post100676 Also, more comments about latency a couple of posts below...
 
Hi Pos,
Thx. Read the Linkwitz link and understand the influence of other crossover points...like you say, all the more reason for linear-phase crossover filters. :)
I'm pretty familiar with phase tracking...been using smaart to align phase traces, subs to main, for a while.
All in all, electrical crossovers are making sense....no real difficulty there..

I'm still not quite so clear with the mechanical part....ie the drivers' raw acoustic response....and corrections for these responses.

I have a hard time seeing how a global correction can correct the phase of each driver individually, through the x-over region. I can see how global can correct the summed response easy enough..
But it still seems like a summed correction is suboptimal to having each of the drivers flattened first..

Sorry for being so dense....
But hey, I really do think I get compensate now :D

I'm kinda trying to wrap my head around the same questions, however if you don't overlap drivers freq range and compensate for the crossover, and then globally linearize the phase over the whole system it should be close to perfect would be my guess. Isn't the combination of driver/cabinet phase shift more important to consider than the drivers phase shift alone?

However, if you linearize the phase on each cabinet/driver individually, could they overlap without phase issues?

Are you supposed to isolate the crossover/cabinet from the spatial room and get it perfect, and use the same settings in every room or is the room part of the equation?

Nice selection of subs you have got there, heard a lot of good stuff about both the OS and the Labhorn. I've used BFM T30's for a few years and a thousand or two hours by now and they are great for what they do but I wanted to make the switch for fewer/more powerfull cabinets... And I can say that 2 keystones I recently built with premium drivers are around twice as loud as 4 almost full-sized T30's with premium drivers, non-scientific opinion offcourse.

Does it take a lot of taps/latency due to the low frequency of the HP on the subs?

Edit: Oh you are building the diy 60° cabinet! I've been following that thread with interest. Awesome system you gotta have there, what music do you generally punish it with?
 
Last edited:
I'm kinda trying to wrap my head around the same questions, however if you don't overlap drivers freq range and compensate for the crossover, and then globally linearize the phase over the whole system it should be close to perfect would be my guess. Isn't the combination of driver/cabinet phase shift more important to consider than the drivers phase shift alone?

However, if you linearize the phase on each cabinet/driver individually, could they overlap without phase issues?

Are you supposed to isolate the crossover/cabinet from the spatial room and get it perfect, and use the same settings in every room or is the room part of the equation?

Nice selection of subs you have got there, heard a lot of good stuff about both the OS and the Labhorn. I've used BFM T30's for a few years and a thousand or two hours by now and they are great for what they do but I wanted to make the switch for fewer/more powerfull cabinets... And I can say that 2 keystones I recently built with premium drivers are around twice as loud as 4 almost full-sized T30's with premium drivers, non-scientific opinion offcourse.

Does it take a lot of taps/latency due to the low frequency of the HP on the subs?

Edit: Oh you are building the diy 60° cabinet! I've been following that thread with interest. Awesome system you gotta have there, what music do you generally punish it with?

I think there is always more overlap that we might guess...for instance, a LR48 (which i view as steep) set at 100hz for both sides, will sum from 60hz to 200hz.

I know you can linearize the filter effect globally.
And I know you can additionally linearize the driver's natural responses individually...still working on this globally :confused:

The plots you saw on my DIY60 link are individually corrected drivers, which definitely overlap.

Room correction? Ugh...I don't believe in it...or shall I say it only works for one position.
Acoustic solutions for acoustic problems...speaker placement..room absortion..diffusion...etc.... these solutions work, but my experience is we are chasing our tail with electronic room correction...

Art's keystones are great huh. I was looking for a project and came close to building a pair before I jumped on the DIY60. And I guess I already had too much sub power.

Yes, numbers of taps (which means latency to me) goes up fast with lower freq.

Music? I play about everything, still have a kid in college, and I'm in my 60's...so everything.
Almost impossible to punish the DIY60...it is one powerful box. It's my poor neighbors that take the punishment,
...or at least until I take it when the cops show up :eek:
 
I have a hard time seeing how a global correction can correct the phase of each driver individually, through the x-over region. I can see how global can correct the summed response easy enough..
But it still seems like a summed correction is suboptimal to having each of the drivers flattened first..

When you have an in-phase complementary minimum-phase acoustical crossover both drivers have the exact same phase shift throughout the crossover.
Once you are there, if you were to correct those drivers to get linear phase acoustical filters for each of them, you would use the exact same correction for both. Hence you can as well put that correction on the input and the result will be identical.
 
When you have an in-phase complementary minimum-phase acoustical crossover both drivers have the exact same phase shift throughout the crossover.
Once you are there, if you were to correct those drivers to get linear phase acoustical filters for each of them, you would use the exact same correction for both. Hence you can as well put that correction on the input and the result will be identical.

Thx for bearing with me POS :eek:

Maybe this will help you see what i'm am missing...

Below are the HF driver (top) and MF (bottom).
Each have a couple of para eqs for a little amplitude smoothing.
They also have linear-phase 48db LRs at 100 & 650, and 650 & 6300.
I know the linear-phase crossovers did not effect phase on either.
Nothing else was done , the corrections (windowing, taps, sample etc) are not what I'm using, just a quick throw together for illustration's sake.

If I look at say 700hz on each, where summation is taking place....
on the HF, I see about 95 deg phase
on the MF, I see about 3.

So about 90 degrees diff, which gives maybe around 3db summation.
OK, so i go measure combined response and easily add para eq to flatten...
But if I adjust phase at 700, what am i adjusting..wouldn't I be moving both drivers the same number of degrees, when i really want to bring them together?
 

Attachments

  • Global Q r.jpg
    Global Q r.jpg
    229 KB · Views: 358
Last edited: