For ABX you should reduce both signals by the same amount of course.
The RMS level should stay the same with or without phase linearization: only peaks level will change.
From the EASY FIR thread :
I don't know what was implemented in Acourate, but to my knowledge any specific amplitude+target curve should result in the same IR (with potentially different windowing shapes and length, of course).
Preringing becomes apparent when you have a linear-phase LP or HP filter (or you linearize the phase of a minimal-phase LP or HP).
The steeper the slope, the more important the prerining becomes.
If you are designing a multiway crossover using convolution, prerinings from matched LP and HP filters (ie amplitude complementary and phase coherent crossover) will cancel each other (as long as the summation is good, so you better match your polar responses and keep your drivers within reasonable distances, especially if you are looking for steep slopes, but that is also true for classical IIR filters, where postringing can become very long...).
If you are also using LP and HP linear-phase filters or linearizing the phase of minimal-phase filters (electrical IIR, or from the box, of natural driver rolloffs...) at the extremities of your loudspeaker frequency range, then you will not have any complementary filter to cancel the preringing, and it might become apparent when looking at the IR.
rePhase is not an automated tool, so you can choose for yourself what to correct and what to leave alone, so for example if you correct the phase of a bassreflex for example you have to know that some preringing will result (audibility of which is another issue...).
Hi soundcheck,
I don't know what was implemented in Acourate, but to my knowledge any specific amplitude+target curve should result in the same IR (with potentially different windowing shapes and length, of course).
Preringing becomes apparent when you have a linear-phase LP or HP filter (or you linearize the phase of a minimal-phase LP or HP).
The steeper the slope, the more important the prerining becomes.
If you are designing a multiway crossover using convolution, prerinings from matched LP and HP filters (ie amplitude complementary and phase coherent crossover) will cancel each other (as long as the summation is good, so you better match your polar responses and keep your drivers within reasonable distances, especially if you are looking for steep slopes, but that is also true for classical IIR filters, where postringing can become very long...).
If you are also using LP and HP linear-phase filters or linearizing the phase of minimal-phase filters (electrical IIR, or from the box, of natural driver rolloffs...) at the extremities of your loudspeaker frequency range, then you will not have any complementary filter to cancel the preringing, and it might become apparent when looking at the IR.
rePhase is not an automated tool, so you can choose for yourself what to correct and what to leave alone, so for example if you correct the phase of a bassreflex for example you have to know that some preringing will result (audibility of which is another issue...).
Hi,
Pre-ringing and post-ringing are typically not an issue.
http://www.bodziosoftware.com.au/Pre_Post_Ringing_IR_And_Pulses.pdf
Best Regards,
Bohdan
Hi there.
Until now I thought the ringing issues are one of the biggest ( and audible) issues associated to FIRs.
DRC also puts some effort on the ringing part. See DRC chapter 4.2.1. + 4.2.2
Cheers
Until now I thought the ringing issues are one of the biggest ( and audible) issues associated to FIRs.
DRC also puts some effort on the ringing part. See DRC chapter 4.2.1. + 4.2.2
Cheers
Hi Bohdan
By "is another issue" I meant "is another question", and a question that does not belong in this thread I think.
It is a general question when dealing with FIR filters. rePhase is just a tool and tries its best not to enforce any practice or dogma by letting the user do what he wants to do, may it result in prominent preringing or not.
Last edited:
It does so by applying special frequency-dependent windowing, and affects the resulting phase correction curves as a result (the fft of the correction impulse), towards a minimum-phase behavior on the ends of the passband.
It has to resort to this technique as it is an automated tool based on impulse inversion.
In rePhase you are in charge of the whole amplitude/phase target, so it is up to you to make these decisions.
To be more precise, if you do not use extremely steep slopes and if you make sure your crossovers topology are sufficiently well behaved (amplitude complementarity around the crossover point, phase coherency, polar response matching, and reasonable distance between drivers relative to the crossover frequency) you will probably never experience any issue (ie, never see any major preringing in the IR of your system, even off axis, let alone actually hear them...).
Having a well behaved crossover topology is paramount for any filtering technique anyway (passive, IIR, FIR...), so it is not really a FIR-related issue. It is just more important with FIR filters, as any target can be obtained...
Having a well behaved crossover topology is paramount for any filtering technique anyway (passive, IIR, FIR...), so it is not really a FIR-related issue. It is just more important with FIR filters, as any target can be obtained...
Hi Jean-Michel,
Such subwoofer was actually tested.
I have done quite extensive testing on linear-phase subwoofer. It was an acoustic 4-th order, with F3dB=16Hz, and LP=200Hz/12dB/oct. Results are in:
http://www.bodziosoftware.com.au/LP_MP_Subwoofer_Tests.pdf
In fact, on this subwoofer, the post ringing was audible, not pre-ringing. Phase linearization removed this problem completely.
If you perform the phase linearization correctly, the bass is tighter, and has more impact and punch (more comments in the paper above). Just to give you an example: the gunshots in the movie "Looper" really hit you hard.
Best Regards,
Bohdan
Hi Soundcheck,
I looked at 4.2.1. + 4.2.2 and then Chapter 4.6.1 “Preventing pre-echo artefacts”.
It says: …..This pre-echo artifacts usually occur on narrow bands and are easily audible as a sort of ringing or garble before transients or sharp attacks In order to avoid them there are basicly two options:
- Reduce the correction on critical frequency regions where pre-echo artifacts may arise.
- Use a minimum phase approach to avoid pre-echoes…..”
This is exactly what I pointed out in my paper http://www.bodziosoftware.com.au/Pre_Post_Ringing_IR_And_Pulses.pdf as Conclusion 4:
“….Pre-ringing is most evident (and audible) while using peaking filters with high gain and high Q-factor, as illustrated on Figure 8 and Figure 24, and suggested by the recording industry. These should be avoided….”
OR - use minimum phase filtering and hope that post-ringing will not be audible.
So, we are all in agreement here.
Best Regards,
Bohdan
Hello Bohdan,
I read your intersting document and was interested by the excerpt :
"Regarding the pre-response. When I came close to the loudspeaker in linear-phase mode, with my ear stacked into the cone, I could hear a short, quite faint noise preceding the main sound. This noise was not there in minimum-phase mode. It is hard to describe this noise, as it was faint, and it did not have any ringing characteristics to it. It was just a short, faint noise."
In fact few people having listened to phase linearized loudspeaker seem to perceived the effect of the pre-ringing as an "intermodulation like" effect when during long excerpts having both a large SPL level of mid and high frequency, a pulsive note having large level and very low frequency content suddenly arrives. In such case, just before the LF pulsive note arrives itself, the mid and high frequency seem suffering some intermodulation by the LF note which is not arrived... (Acausality problem).
This should explain why the attacks of such LF pulsive notes seem somewhat unnatural...
For what its worth...
( Edit : Sorry I sent this message before I received your message saying how to cure that phenomeneon...
)Jean-Michel Le Cléac'h
Last edited:
Hi Jean-Michel,
Sorry, I missed your post.
Again, I would like to keep these general discussions out of that thread if possible: rePhase is a general tool that let the user design any amplitude/phase target he wants (linear/minimal/mixed/whatever phase). Preringing, postringing and others "goodies" are just a consequence of a given target, and their audibility is the responsibility of the user when choosing (and testing) these targets.
Good practices exist, and as with any tool it is more a matter of how you use it than the quality of the tool itself: you can design an minimal-phase filter with rephase, that will give the exact same result as an IIR would have (albeit without rounding errors when applying the correction, vs typicall biquads implementations...), you can design crossovers with steep minimal-phase slopes, 100dB peaks with a Q of 40, or do whatever you want, and accept the consequences
Regarding phase linearization at the ends of the passband (where no complementary filter will occur to cancel ringings), it is just a trade of post ringings for preringings.
Here is an illustration with the linearization of a LP filter (easier to show on the impulse, but the situation would be similar with a HP filter):
The total energy of the ringing (and intermodulation it could provoke, to follow your argument) is identical, but it becomes symmetrical when phase is linearized: total energy *is* the same.
You will have to try for yourself if that is audible or not in your situation (audibility of preringing *and* audibility of phase linearization
)Beside my own experience, reports I have read from users of rephase claim the most audible gain when linearizing crossovers in the 80Hz-300Hz range (and that is in small room in most cases, so around and under the Schroeder frequency).
Please try for yourself: it is quite easy using a convolution plugin in your mediaplayer like foobar+convolver. I would certainly be interested in hearing your opinion
(but I think you do not have a midbass or subwoofer in your installation, so that point is probably moot).The fact that you cannot easily measure phase at the listening position (because of room issues) does not necessarily mean there is no gain (it just mean the measurement technique is not up to the task: averaging in the amplitude domain gives good results, but of course that is not as easy to do when dealing with phase measurement, but could be done anyway...).
Having a more compact IR is a good thing anyway, and that can be seen even at the listening position (in a decent room...) by just looking at the measured IR.
I have no firm opinion on the bassreflex linearization itself, but if people hear timing differences between sealed (12dB/oct, 180° phase shift) and BR boxes (24dB/oct, 360° phase shift) then there is obviously something there...
Last edited:
Hi folks.
@Pol.
My intention is not to highack your thread. I'm just wondering how to use
your tool to achieve best possible results.
If you look at Acourate and its user group, you'll see that a huge effort is taken to explain properly and discuss how to create your filters. This is
a more then challenging task, with endless options (and traps), for the average user.
One more question.
Avantgarde Acoustics introduced an active speaker line called Zero 1.
They use FIRs inside. What made me wonder was a term called
"progressive Firs". Basically a pretty flat (6db) filter in the beginning, which gets steeper on its way.
Link
I'm wondering if this relates somehow to above discussion.
Cheers
@Pol.
My intention is not to highack your thread. I'm just wondering how to use
your tool to achieve best possible results.
If you look at Acourate and its user group, you'll see that a huge effort is taken to explain properly and discuss how to create your filters. This is
a more then challenging task, with endless options (and traps), for the average user.
One more question.
Avantgarde Acoustics introduced an active speaker line called Zero 1.
They use FIRs inside. What made me wonder was a term called
"progressive Firs". Basically a pretty flat (6db) filter in the beginning, which gets steeper on its way.
Link
I'm wondering if this relates somehow to above discussion.
Cheers
Hi
It was not directed at you at all (heck, I am the one how quoted your question from the EASY FIR thread here
)You can experiment with special slopes in rephase, like the "reject low" and "reject high" slopes.
You would have to load the impulse in a visualization software (like cool edit, or HOLM if you want to see the effect of its convolution on your measured response) to see the effect on the ringings this will have.
I think the most important thing to remember when building your crossover is to always consider the whole acoustical filter response (electrical+driver+box+...) and always aim for complementarity in amplitude and coherency in phase throughout the crossover (1 octave above and below, more or less, depending on the slopes...).
These practices do not only apply to FIR filters (IIR and passive filters should also be build that way), but become even more important when the user can design arbitrary amplitude and phase responses (brickwall and the like...).
I agree. When I was working with Bohdan on the UE and testing it the one area where I noticed the most difference was in the bass response. Phase linearization at higher frequencies seemed apparent to me only with certain typed of source material. Also, one of the things I did was to measure the impulse response of a system with linear phase and compare it to one with typical LR4 phase response on and off axis horizontally and vertically. Note that the test pulse was filtered to give it an exponential tail more characteristic of a transient decay. In a sense, I consider the tweeter response of the conventional speaker to be a form of pre-ringing since it precedes the woofer response.
An externally hosted image should be here but it was not working when we last tested it.
Hello Bohdan,
I read your intersting document and was interested by the excerpt :
"Regarding the pre-response. When I came close to the loudspeaker in linear-phase mode, with my ear stacked into the cone, I could hear a short, quite faint noise preceding the main sound. This noise was not there in minimum-phase mode. It is hard to describe this noise, as it was faint, and it did not have any ringing characteristics to it. It was just a short, faint noise."
In fact few people having listened to phase linearized loudspeaker seem to perceived the effect of the pre-ringing as an "intermodulation like" effect when during long excerpts having both a large SPL level of mid and high frequency, a pulsive note having large level and very low frequency content suddenly arrives. In such case, just before the LF pulsive note arrives itself, the mid and high frequency seem suffering some intermodulation by the LF note which is not arrived... (Acausality problem).
This should explain why the attacks of such LF pulsive notes seem somewhat unnatural...
For what its worth...
( Edit : Sorry I sent this message before I received your message saying how to cure that phenomeneon...
Jean-Michel Le Cléac'h
Hi Jean-Michel,
I think you are reading way too much into my comment.
I have reported things as accurately as I could hear them. With my ear stacked into the cone, I could hear down to -80dB level or lower. Only then I could perceive a short and still faint noise. As I mentioned further down in my paper, I did not have any explanations for this issue, and it was faint noise, not pre-ringing.
There is absolutely no way you could hear this in normal listening position, on any musical material.
“….I would expect the pre-response to resemble a sound rather than a noise, because this is the way it manifests itself in the impulse response of the brick-wall filters – as a pre-ringing. I do not have an explanation for my observation, and since it was not audible at normal listening distance, I mention it here for the record only. Listening tests conducted on phase-linear loudspeaker (Preprint 2927) by Mr Richard Greenfield, Dr Malcolm Hawksford and several other expert listeners also did not reveal any pre-response issues. Concluding from all the above, the pre-response of stand alone subwoofer (-12dB/oct, 120Hz Linkwitz filter), in normal HT application is not an issue….”
Please provide the reference pointer to your comment. I should be able to repeat your tests and I will be happy to report on the outcome.
Best Regards,
Bohdan