 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
207.216.246.51
...and if you are, what are you using?
Thuneau Phase Arbitrator
BruteFir/Acourate
Convolver in Foobar/Winamp
Others...
I have set up Thuneau's Phase Arbitrator to get much improved impulse responses. "Room correction" methods like BruteFir and impulse response convolvers affect both time and amplitude domains - so they're not really just phase correction but "frequency response" correction, where frequency response REALLY means amplitude AND phase response. I have also meddled with convolver methods, but WOW the learning curve is steep and there are many variables that make the end result a little too... "subjective" for my liking. I can go through the PROCESS, but have not had the time to experiment with different variables such as bkfile curves and the like.
Cheers,
Presto
Follow Ups:
Just last week I spent an afternoon on a DRC proof of concept using Inguz EQ on my Squeezebox. I installed all of the necessary software and borrowed an AT2025 mic and M-Audio MobilePre to make the measurements.
After generating and applying the Inguz/DRC filters, I am fairly impressed with the apparent improvement in sound.
My next step is to purchase a more appropriate measurement mic (ECM8000) and make some better room measurements using REW or equivalent RTA software. I want to measure the response before and after DRC to *see* and *confirm* what my ears are hearing. I also want to see if I can improve the response before DRC by moving my speakers around and rearranging things (don't have much flexibility here).
I find this stuff very interesting.
-Matt
Of late I have been using IK Multimedia's ARC, a room correction VST plugin which deals with amplitude and phase (to a certain degree). I wrote a little review of it in this forum a few weeks back. It is only customizable to a slight degree, but so far I have found the results to be very good in my room because it doesn't seem to overcorrect. Much easier to implement than DRC via Brutefir. In this case I think the lack of options and the idiot-proof measurement technique helps obtain good results rather quickly.
Alan
Alan:
Right on - another toy to play with!!
I am going to look into this and see if the demo actually permits a listening test of sorts.
"Overcorrecting" is a big issue with DRC - at least it was for me anyways. How much correction to apply (and where to apply it) is the subjective part of DRC - that's what I think about it anyways...
CHeers,
Presto
.
only phase coherent over a portion of the audio bandwidth, then you don't need phase correction!
Any speaker that uses conventional drivers will have a tweeter low-end rolloff and a midbass high-end rolloff. Unless both drivers roll off at 6db / octave (1st order acoustic) and NO electrical filters are used, then the design will have 2nd order or greater slopes in there SOMEWHERE.
Simple tests can reveal at which frequencies the speaker is phase coherent. But when drivers are rolling off, it's a safe bet that higher order slopes are evident, making phase coherence not at all possible.
THere are only two transient perfect speakers that I am aware of in the passive domain - subtractive delay and filler driver types.
1st order "phase coherent" are not included in that family, but could be considered a close second. A fullrange driver is also not phase coherent, since it also has both a low end and high end rolloff making it behave more like a midrange with a 12db/octave acoustic lowpass and highpass.
Phase correction can eliminate phase shifts completely for the entire bandwidth - even the low frequency rolloff of woofers and phase shifts that occur with ported bass enclosure designs. THat is what makes it so... interesting.
Cheers,
Presto
The "phase response" of digital playback, especially when using linear-phase filtering (which is utilized in over 90 percent of playback), is already exemplary...... Any "correction" wouldn't really do anything..........
Now if it's to correct the phase response of a loudspeaker, that would be a different story. But the designer of the correction must know the phase response of your particular loudspeaker in order for it to work......
For one point in space.
This is why BruteFIR is difficult to configure. It requires you to take measurements of your speakers in your room, then constructs a FIR convolution filter that is used to transcode the playback of your source material in perfect acoustic phase (for one point in space).
Btw, the jury is still out how important "perfect" acoustic phase response is. S. Linkwitz has an interesting test on his website that is worth trying out.
Any technique that uses a measurement at the "sweet spot" for a reference, yes, this is true. The more correction that is used, the more "location dependent" the system becomes.
But for the Thuneau Arbitrator, no such measurements are necessary for it's operation. Granted you may need to measure to find out the system transfer functions to dial in the Arbitrator properly, but the Arbitrator does not make use of these measurements itself.
As such, the Arbitrators phase correction is valid for as many listening positions as the speaker was originally capable of. I think this is what makes it stand out from the other correction methods available.
Cheers,
Presto
Add the DEQX PDC to your list. It uses FIR filters like BruteFir/(((acourate))). The main advantage of the PDC is that it's *very easy* to setup and configure.
(Note that there is also a BruteFir plugin available for the Squeezebox.)
How do you know if your music or track is out of phase or needs correcting?
Or do you just go back and forth to listen for better/more preferred sound?
Music is the Bridge between Heaven and Earth - 音楽は天国と地球のかけ橋
You're talking about absolute polarity, which is simply a flip of the ENTIRE waveform. It is uniform across the frequency spectrum, in that all data is essentially multiplied by (-1) regardless of amplitude or phase.
Phase roll is caused by speaker crossover slopes higher than 1st order. 2nd order causes a 180 degree phase shift. 3rd, 270. 4th, 360. Some believe this phase shift only (magically) occurs AT the crossover point but it does not. The phase shift affects all frequencies in the passband of the crossover. Phase shifts cause delays, because phase shifts occur in the time domain. The first order crossover has the unique property of having what is called a constant group delay. This means the phase-shift-induced delay is NOT frequency dependent but constant, thus time alignment (another source of delay) can be used to correct for this. This is done by aligning the acoustic centers of the drivers with respect to a common plane (on a flat baffle) or using a digital delay when using digital crossovers. All other orders higher than 1st order (2nd, 3rd 4th, etc.) have increasingly "worse" group delay.
Although a 4th order crossover has a full 360 degree shift and requires no drivers to be connected in reverse polarity is it often revered as being a crossover which is "phase coherent". Sure it is. But only at the crossover point. It suffers from the same group delay issues as a 2nd or 3rd order (and is in fact worse because of the higher 4th order slope). The greater the group delay, the more square wave reproduction is distorted - aka the further you get from transient accurate reproduction.
If you pass waveforms of MUSIC through a transient perfect crossover, they can pass through unchanged. But a non-transient perfect crossover - even the holy grail 4th order - will change the waveform so it is almost unrecognizable. Things still SOUND very much the same but they are not the same. If you take the harmonic structure of specific instruments for example, these signatures have both amplitude and time domain characteristics. By mucking with the time domain aspects, you are no longer getting the SAME waveform. It sounds very close, but it's not the same. TP (Transient perfect) crossovers also will sum the outputs to pass a square wave. People often say "But I don't listen to square waves!" which is true, but square waves are the defacto standard test for transient perfect reproduction. If they say this, they need to know that if a crossover is mucking up a square wave (at ANY frequency) then part or all of the crossover is mucking up phase response and, as a result, the delicate sonic signatures of every instrument and voice and sound passed through it.
As for general audibility of phase distortion, the naysayers say "Well, phase distortion is only audible under very controlled conditions in ABX comparisons with SPECIFIC test tones". So the same people who argue that they don't listen to square waves use the waveshape argument to attempt to debunk the validity of accurate phase reproduction. These people claim that ABX tests don't work on cables because you need to listen for weeks at a time, yet they won't concede to doing this for phase distortion tests, making them somewhat hypocritical at times. To answer to this, I normally respond "Yes, but I listen to music - not waveshapes that demonstrate phase distortion most obviously".
The improvements from having transient perfect reproduction are very subtle, but I believe them to be very real. Sounds seem more "right". It's cliche, but that's what it is. Instrument timbres seem to take on a very real and lifelike and "palpable" character. Shrillness of the flute and some string instruments seems to vanish, while the soul of these instruments seems to come out. And spatially, where speakers vanish and create a convinving sound stage with TP speaker systems the walls seem to vanish as well. You get closer to being taken out of your listening room somewhere else - to a concert hall, a jazz club or recording soundstage. This, for me, is why I BECAME an audiophile - to try and get closer to this experience. Forget width and depth. You feel more like you are THERE.
Believe in it's audibility or not, the arbitrator does exactlty what it is advertised to do. I've proven this for myself with exhaustive testing with both test signals and music material. I've experimented with 2nd and 4th order LR crossovers, 3rd order BW and 4th order Deulund variants (which require cascading of two Allocators in a VST shell and a thorough understanding of filter coefficients - not easy but I eventually did it).
Duelund worked okay but was really not worth the effort when using phase arbitration. A Deulund crossover is very close to a Linkwitz Riley crossover of half the order. I.e., a 2nd order Duelung approximates a LR4 and a 4th order Duelund approximates a LR8.
I don't know why but although the Arbitrator improves all crossover functions very well, it seems (to me anyways) to give PERFECT results with the LR4. Could be something to do with the math involved in a perfect 360degree phase shift at the crossover point, I am not sure. I don't have the algorithm to hand to a mathmetician for an algorithm and truly I don't really care - the Arbitrator works.
It works when using just the filters mathmatically and it works in the real world and I have *measured* before/after impulse responses to prove it (to myself anways).
Cheers,
Presto
Maybe just semantics, but the most commonly used 4th order these days is probably the Linkwitz Riley alignment, and it is phase-coherent throughout the frequency spectrum, meaning that if you check the driver outputs at any frequency, not just at the crossover frequency, they will be in phase.
But as you say, the 4th order crossover does have more group delay, so the high frequencies would ultimately be rotated by a full 360 deg in reference to the low frequencies. It is modeled as an all-pass filter. The summed outputs will have constant amplitude throughout the frequency spectrum since they are always in phase with each other, but not with the input except at very low frequencies.
But as you pointed out the 4th LR order designs are not linear phase. "In phase at all frequencies" is necessary but so is a constant group delay.
In simpler terms, LR4 is not transient perfect despite being phase coherent. Only 1st order acoustic designs (sort of) are, as well as substractive delay, filler driver and phase corrected designs.
It's not semantics, you are correct. I should have said "transient perfect" not "phase coherent". LR4 is indeed a special case, despite not being transient perfect. It is, however, the easiest to correct in my (rather limited) phase correction experience. Could be the phase-coherent aspect being beneficial. LR4 has no lobing error as well - thus it's popularity with the prosound folks.
CHeers,
PResto
hey presto,
I´ve been using the allocator/arbitrator for almost a year now and I must honestly say it works very well indeed. -as per se !
....but I must confess that setting the right arbitrator / phase responses is a very hard task on audition trial-and-error !
-maybe my magnepan dipole system is not bettering the cloud, but I´m working on it.
there are just an enormous amount of parameters to take into account.
I´m not sure if I can call arbitrator it a definite gain, before I´ve measured and adjusted the entire system for phase errors. -and thats inevitable gonna be with square waves and the lot....
-and that´s got to be my audiophile challenge for 2010.
if you have any suggestions to achieve such analysis , I´m very happy to hear about it !
kindest regards
Playmate:
You need a measurement mic and small mixer with phantom (48V) power. I use a Behringer ECM8000 mic and a Behringer Eurorack UB802 mixer. For software, you need a MLS suite - something that will capture, at a minimum, the amplitude response of your system. You could record a sweep with a trial version of Adobe Audition, and then analyse the amplitude response. Or, you could try to get a copy of ETF which has limited but useful measurement capabilities in demo mode.
You will need to run sweeps of the system of each set of drivers individuall. In a two way, the midbasses by themselves, and tweets by themselves, one speakers at a time. For a three way, woofs, then mids, then tweets. A full range sweep will tell you little about where the crossover point is and what it's slopes are.
Then, you need to import the data as FRD files into Allocator and dial in the crossover point(s) that match up best with where the measured curves intersect. You also want to determing the slope of the curves. Basically, you want to enable an arbitrator curve for each crossover point that lays on top of what you have measured.
Manufacturers data for drivers is of no use. This will tell you nothing about the response on the specific baffles you are dealing with and what electrical filtering has been done.
The only time you could get into trouble is if the desiger chose to use asymmetrical slopes. If there are asymmetrical slopes, you may be in luck as the designer may have implemented a subtractive delay design and you are already transient perfect. But I would bet very few designers are using this method.
Cheers,
Presto
Thanks Presto.
Your explanation obviously matches what Allocator designer told me.
It is also possible to use ARTA free software to measure.
Of course both Playmate and I have speakers with asymmetrical slopes :-/
Bibo:
How is your system coming along? Sorry I have not replied to all of your e-mails. I have been extremely busy with work, 2 kids in diapers and a wife who's been not exactly well since baby 2 came along.
I was wondering if you ever did get Allocator to work with your AV system and if the delay caused by the Allocator process was noticeable? (Are there timing issues with lip movement and speech for example).
Anyways - yeah I would ask Jan on the Allocator forum what one can do about asymetrical slopes. The only way to go about things now is to add a "pair" of matched curves around a single Fc point AFAIK. Would be nice if each individual curve had selectable slop. Perhaps this does not agree with the math involved in arbitration.... who can say. On Jan would know for sure!
So, Merry Christmas to you and yours!
Cheers,
Presto
Re. my system, I still would need your help to troubleshoot. It seems that in VAC -> VST Host -> Allocator -> Asio4All configuration, ASIO4All creates channel swapping and crackling during multi-ch playback. I can't really use it as it is. I have DS multi-ch playback with no XO and let bass and mid/high receive full signal.
When you have a chance, I will still gladly accept your help. In meantime, I wish the best to your wife.
Re. phase with Allocator/Arbitator, I asked Jan about asymmetrical slopes measuring. He replied:
"The best way would be to plug in a mic and tweak in real time using a measurement program. There are a few out there including the 30 day demo of Smaart. You can even measure using real music as source.
You would have Arbitrator in line and tweak it until flattest phase response is found.
Simply modeling, is kind of flying blind and relying on measurements taken by other people of their speakers. Yours could be off by as much as 5-10% either way from the published curves.
I know it's not easy, but very satisfying when accomplished."
After you adjust the delay in the signal to the drivers you will get equal delay only when the listener is equidistant between the two drivers. If the listener is elsewhere then there will be frequency dependent interference, a.k.a. comb filtering. A small digital delay isn't likely to be much different than moving your ears up or down a few inches (assuming the two drivers are in a vertical line). That's not to say that it may not be more convenient to tweak some filter coefficients than tilting speaker cabinets up and down (or adjusting the number of books one sits on...).
As to absolute polarity, it will be difficult if not impossible to hear on a speaker that isn't phase coherent. On the other hand it may be easy to hear on a single driver table radio, despite its lack of frequency extension due to the single driver and lack of crossover.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Tony
You said very well: "As to absolute polarity, it will be difficult if not impossible to hear on a speaker that isn't phase coherent. On the other hand it may be easy to hear on a single driver table radio, despite its lack of frequency extension due to the single driver and lack of crossover."
Agree 100%!
But would add - On well recorded music, correct absolute phase vs incorrect absolute phase is very easy to hear when the speaker is phase coherent.
Frank
Thanks this is good information. I am going to have to try it out, I think I found the foo_dsp_vst plug in so I can start experimenting.
Music is the Bridge between Heaven and Earth - 音楽は天国と地球のかけ橋
Post a Followup:
| FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
||||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2024, Audio Asylum, all rights reserved ]
