DiyMobileAudio.com Car Stereo Forum banner

1 - 20 of 115 Posts

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #1
I am working on a new system, and I am a fan of loudspeakers that keep the drivers in phase.
There is only one type of filter that can do this: a first order filter.
This is why first order filters are used by Vandersteen, Dunlavy, Thiel, Dynaudio, and numerous others.

I do not think that it is a coincidence that those speakers are known for their startling imaging and wide soundstages; to my ears the biggest improvement rendered by drivers that are in-phase is the soundstage. (My home speakers are Vandersteen and I use first order filters almost exclusively in my projects.)

There are a couple designers who use high order crossovers that are in-phase, or close to it. Tom Danley is one, Jean Michael LeCleach is another.

I've been avoiding trying to learn how to do this, because phase makes my head hurt, but I decided it's about time to learn.

I'm not 100% sure if the two designers are using the same method, in fact one of the reasons I'm posting this is to get an understanding of how it works.

Some caveats:

1) The material in this thread is Jean Michael LeCleach's
2) I don't see a copyright on it, but if I missed it, please let me know. I'll take down the images and link directly to the source.
3) The reason that I don't just post a link to the source is that Mr LeCleach passed away last year, and I've seen many times when good solid information fell into obscurity because the author died. So this thread is definitely an attempt to keep this method in the limelight.

 

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #2

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #3






 

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #4





 

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #5





 

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #6










More posts to come, there's about fifty more slides in the presentation. (Entire file available from link in second post.)

 

·
Registered
Joined
·
5,098 Posts
Good to have you back, hope you had fun :). Reading and picking up on some things.

Phase is chaotic in reverberant environments, yes, but the direct sound is not affected by reverberation. In some situations such as choral music in a cathedral from the back of the audience, phase is totally messed up, but in most other cases it still matters!
So, the relevance of phase depends on the environment. Not only are you at the back of the cathedral, the roof is 6" above your head and the furthest wall is like 3 feet away. The amt of direct energy you're hearing is WAY less than what you would in a room. It is a phase nightmare. Phase matters in a room but in a car.....

Three experiments were performed which confirm the audibility of time offset in loudspeaker drivers but indicate that this audibility is due only to the frequency response aberrations resulting from the time offset. Implications of these results are discussed.
You're hearing phase anomalies in the response domain. Response is the way one hears phase.

Given that the response alone can explain time offset audibility and time delay alone cannot explain this audibility it seems inescapable to conclude that arrival time compensation by itself has no audible value.
Again room vs room. Timing is of vital importance in a car as it is the only way to make the speakers equi-distance from you. Something that is taken for granted in a room. I think in a car getting the timing right and then smoothing out the phase by using the eq is the way to go. IMHO eq around the xover point is basically to ensure a smooth transition from one set of drivers to the next.

"A frequent argument to justify why phase distortion is insignificant for material recorded and/or reproduced in a reverberant environment is that reflections cause gross, position sensitive phase distortion themselves. "
"Although this is true, it is also true that the first-arrival direct sound is not subject to these distortions
Environment. In a car we are getting precious little of that direct sound. Not only is the quantum of direct sound vs reflected radically different between a room and a car, the time interval between the direct and the first reflection (gating?) is much smaller in a car.

4th order LR crossover have always sounded "disjointed" to me - transients sound blurred, and high frequencies don't match up with low frequencies.
In a room, speakers that use shallow slopes image, stage great and are tonally decent in a small spot X. Creating and staying in that small X is just not possible in a car.

In a room outside that X, the stage and image collapse a bit and acoustical interference? makes the speakers sound 'darker' in the mid range. B&O is a bad example to use because they are playing with timing on most of their speakers, iirc.

Attributes like space, air, etc in a car, is better achieved on the back of wider dispersion on steeper slopes vs narrow dispersion on shallow slopes.

It's beginning to get too technical for me.:)
 

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #8
I've never heard a speaker with high order slopes that could image as well as a speaker with first order slopes.

Getting the phase flat from 250hz to 20khz would be tricky; it basically requires us to eliminate all reflections within almost half a meter(!)
But if you limit yourself to 1khz-20khz, it's pretty easy to do with first order slopes.

I think flat phase is probably the main reason that Hybrid wins so many prizes, their designs use high order xovers but their xover points maintain flat phase in the midrange.

LeCleach offers a tantalizing possibility here, the ability to use high order slopes *and* have (close to) flat phase.

Another bonus is that it requires the midrange to be closer than the tweeter, which is exactly how most cars are set up.

 

·
Moderator
Joined
·
18,358 Posts
the bit about low order crossovers being better transiently is true to some extent (various types change) BUT only in the absolute term of the crossover itself. When you pair the crossover with a driver things change. When you pair a driver+crossover with another driver+crossover, things change. A 1st order BW (with 0ms GD) added to driver paired with another crossover will alter the transient (GD) and can just as easily result in as high GD as using a higher order crossover slope.

The summed response (frequency, phase/polarity, time) is the important aspect and evaluating a single crossover type on its own is only a piece of the puzzle. Don't get caught up the crossover types themselves. It's important to know what they do, but the interaction with the driver you're using it on and other crossovers in the network is what matters.

One cool thing is in the car we have the luxury of DSP. Practically everyone on this site has DSP. Heck, I'd venture to say that more people in car audio use T/A than those in home audio do. Therefore, if we set a crossover network to foremost limit the driver's bandwidth (for reasons such as beaming, distortion, or breakup) then you have the luxury of adjusting time of one driver relative to the other to lessen GD IF you feel that's of utmost importance. Maybe between a tweeter/mid at higher frequency this isn't as much a concern as it is with low frequency crossovers where the wavelength is so long that phase is more impactful on group delay.

For example, if you had a midrange that needed to be low passed at 3khz with a steep filter and the tweeter network you have resulted in you having a 90deg phase shift at the summed crossover point of, say, 2.5khz, you could delay the tweeter by 0.10ms and achieve a 90deg phase transformation that would *theoretically, in this hypothetical case* get you in phase at the crossover.
 

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #12
the bit about low order crossovers being better transiently is true to some extent (various types change) BUT only in the absolute term of the crossover itself. When you pair the crossover with a driver things change. When you pair a driver+crossover with another driver+crossover, things change. A 1st order BW (with 0ms GD) added to driver paired with another crossover will alter the transient (GD) and can just as easily result in as high GD as using a higher order crossover slope.

The summed response (frequency, phase/polarity, time) is the important aspect and evaluating a single crossover type on its own is only a piece of the puzzle. Don't get caught up the crossover types themselves. It's important to know what they do, but the interaction with the driver you're using it on and other crossovers in the network is what matters.

One cool thing is in the car we have the luxury of DSP. Practically everyone on this site has DSP. Heck, I'd venture to say that more people in car audio use T/A than those in home audio do. Therefore, if we set a crossover network to foremost limit the driver's bandwidth (for reasons such as beaming, distortion, or breakup) then you have the luxury of adjusting time of one driver relative to the other to lessen GD IF you feel that's of utmost importance. Maybe between a tweeter/mid at higher frequency this isn't as much a concern as it is with low frequency crossovers where the wavelength is so long that phase is more impactful on group delay.

For example, if you had a midrange that needed to be low passed at 3khz with a steep filter and the tweeter network you have resulted in you having a 90deg phase shift at the summed crossover point of, say, 2.5khz, you could delay the tweeter by 0.10ms and achieve a 90deg phase transformation that would *theoretically, in this hypothetical case* get you in phase at the crossover.
Yeah phase response is a skull crusher. Here's an example:

In my Synergy horns I use first order electrical crossovers nearly 100% of the time. Occasionally I find that I have to flip the polarity of one driver to get the phase flat. This is always kind of a bummer, because theoretically you can't get a midrange and tweeter in-phase with a first order slope UNLESS the drivers are both connected in positive polarity. But my microphone doesn't lie, and many times I have to flip the polarity to get a phase response that measures flat.

But now I can see what's going on:

What's happening is that the combination of the natural rolloff of the tweeter and the electrical rolloff of the filter is combining to create a slope that's more like third order, not first order. Combine that with the fact that the midrange is about 10cm in front of the tweeter, and I accidentally had a configuration that was pretty close to what LeCleach is recommending in this paper.

This is way cool, because it's very difficult to get a "true" first order rolloff with a tweeter. Tweeters typically have a 'Q' that's between 1.0 and 2.0, and due to that, they tend to rolloff steeply. If I'm not mistaken, this is one of the reasons that the Dynaudio, Thiel and Vandersteen crossovers are so insanely complex; they use a pile of components to create a "true" first order slope.

 

·
Registered
Joined
·
1,912 Posts
For example, if you had a midrange that needed to be low passed at 3khz with a steep filter and the tweeter network you have resulted in you having a 90deg phase shift at the summed crossover point of, say, 2.5khz, you could delay the tweeter by 0.10ms and achieve a 90deg phase transformation that would *theoretically, in this hypothetical case* get you in phase at the crossover.
This is something I am learning about, would the .1ms delay affect the rest of the tweeter's phase? IE everything above the xover point...
 

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #14
This is something I am learning about, would the .1ms delay affect the rest of the tweeter's phase? IE everything above the xover point...
What he recommended will work... at the crossover point.

That's the problem with high order crossovers: the phase shifts dramatically and the phase shift changes with frequency. So if you fix it at one frequency, you may well make it worse at another.

The higher the order, the more the shift. First order has 90 degrees of phase rotation, 2nd is 180 degrees, 4th is 360 degrees, etc.

If you really want to make your head explode, there is a piece of software named 'RePhase' that can apply delay that varies with frequency... IE you can basically flatten the phase of any filter! This is some Next Level Shit, and it used to require a whole rack of processing and about $100,000 in gear, but RePhase is free.

 

·
Registered
Joined
·
1,912 Posts


What he recommended will work... at the crossover point.

That's the problem with high order crossovers: the phase shifts dramatically and the phase shift changes with frequency. So if you fix it at one frequency, you may well make it worse at another.

The higher the order, the more the shift. First order has 90 degrees of phase rotation, 2nd is 180 degrees, 4th is 360 degrees, etc.

If you really want to make your head explode, there is a piece of software named 'RePhase' that can apply delay that varies with frequency... IE you can basically flatten the phase of any filter! This is some Next Level Shit, and it used to require a whole rack of processing and about $100,000 in gear, but RePhase is free.



Funny enough, I just learned about RePhase a couple weeks ago...definitely planning to use it with my home 2 channel build.
 

·
Premium Member
Joined
·
5,972 Posts
Discussion Starter #17
Funny enough, I just learned about RePhase a couple weeks ago...definitely planning to use it with my home 2 channel build.
I was babbling about Synergy Horns over at diyaudio* and the inventor was kind enough to post a pic of what RePhase can do to his invention:



This shows the frequency response, impulse response, and phase response.
Check out the impulse response in particular, see how it 'fades to black' almost instantaneously? This is something that's literally impossible with the LR4 filters that are so popular. Basically the proaudio guys mostly don't care about phase, but they care a LOT about power handling, so LR4 is the pro audio default. IMHO AudioControl made it popular in the car audio world, and AudioControl is a few miles from Rane, which is a pro audio company. (I used to live a few miles away from both of them.)

TLDR: I think LR4 filters are used WAY too much, and I think AudioControl was influenced by Rane, due to their physical proximity.


If you Google "perfect impulse response", this is the pic that comes up. It's a Dunlavy SC-IV which uses, you guessed it, first order filters. Both Dunlavy and LeCleach have passed away, so hopefully I can keep some of these ideas out there.

* original discussion, with Geddes, Danley and myself here : http://www.diyaudio.com/forums/multi-way/178187-great-waveguide-list-22.html#post3912653


Also, that phase response is just completely ridiculous too. Basically flat from 100hz to 13,500hz. The only reason there's a blip at 13,500hz is because that's the end of the horn! (13,500hz = 1")
 

·
Registered
Joined
·
5,098 Posts


What he recommended will work... at the crossover point.

That's the problem with high order crossovers: the phase shifts dramatically and the phase shift changes with frequency. So if you fix it at one frequency, you may well make it worse at another.

The higher the order, the more the shift. First order has 90 degrees of phase rotation, 2nd is 180 degrees, 4th is 360 degrees, etc.


iirc, with both a LP and HP the phase shift occurs in frequencies below the xover point. Now, with the same xover point and the same order for both LP and HP, the phase shift on the both drivers is identical.

We hear phase anomalies in the response domain. Therefore in a car that is where they should be corrected along with timing. In a car you need timing and response to correct for any 'phassy' issues Also, any significant phase issue will more likely be heard on the driver with the LP simply because the anomalies are occurring in the drivers pass band. With the HP driver, these anomalies occur beyond the pass band and are rolling off steeply on a fourth order slope. With fourth order slopes, I will eq each driver an octave above and below the xover point for a smooth combined response in this region as measured and finally as heard. Getting these zones right is finally what gives that sense of flowing and connected unity of sound and source.

This range of an octave above and below, could fall in any zone of your overall response curve, i.e. one that is flat, sloping down or up. The transition has to be smooth the range itself does not have to be flat.

Additionally since it's a car, the environment and install ensure that a 4th order electronic slope does not normally translate into a 4th order measured / acoustic slope. So you need to eq for that as well.

You don't need to do any of the above in a room. In a car nearly everything, comes down to response one way or the other, like Erin mentioned. A good sounding car is almost always about micromanaging the perceived response.
 

·
Registered
Joined
·
1,912 Posts


I was babbling about Synergy Horns over at diyaudio* and the inventor was kind enough to post a pic of what RePhase can do to his invention:



This shows the frequency response, impulse response, and phase response.
Check out the impulse response in particular, see how it 'fades to black' almost instantaneously? This is something that's literally impossible with the LR4 filters that are so popular. Basically the proaudio guys mostly don't care about phase, but they care a LOT about power handling, so LR4 is the pro audio default. IMHO AudioControl made it popular in the car audio world, and AudioControl is a few miles from Rane, which is a pro audio company. (I used to live a few miles away from both of them.)

TLDR: I think LR4 filters are used WAY too much, and I think AudioControl was influenced by Rane, due to their physical proximity.


If you Google "perfect impulse response", this is the pic that comes up. It's a Dunlavy SC-IV which uses, you guessed it, first order filters. Both Dunlavy and LeCleach have passed away, so hopefully I can keep some of these ideas out there.

* original discussion, with Geddes, Danley and myself here : Great Waveguide List - Page 22 - diyAudio


Also, that phase response is just completely ridiculous too. Basically flat from 100hz to 13,500hz. The only reason there's a blip at 13,500hz is because that's the end of the horn! (13,500hz = 1")
Reading this gave me goose bumps. Can't wait to try it next week when I get my computer system set up. Your Rane/AC conjecture is totally plausible. Employee cross pollenation is certainly a cause for this trend. RePhase lets us cross steep but retain power handling, right? We get our cake and eat it too?

iirc, with both a LP and HP the phase shift occurs in frequencies below the xover point. Now, with the same xover point and the same order for both LP and HP, the phase shift on the both drivers is identical.

We hear phase anomalies in the response domain. Therefore in a car that is where they should be corrected along with timing. In a car you need timing and response to correct for any 'phassy' issues Also, any significant phase issue will more likely be heard on the driver with the LP simply because the anomalies are occurring in the drivers pass band. With the HP driver, these anomalies occur beyond the pass band and are rolling off steeply on a fourth order slope. With fourth order slopes, I will eq each driver an octave above and below the xover point for a smooth combined response in this region as measured and finally as heard. Getting these zones right is finally what gives that sense of flowing and connected unity of sound and source.

This range of an octave above and below, could fall in any zone of your overall response curve, i.e. one that is flat, sloping down or up. The transition has to be smooth the range itself does not have to be flat.

Additionally since it's a car, the environment and install ensure that a 4th order electronic slope does not normally translate into a 4th order measured / acoustic slope. So you need to eq for that as well.

You don't need to do any of the above in a room. In a car nearly everything, comes down to response one way or the other, like Erin mentioned. A good sounding car is almost always about micromanaging the perceived response.
This is how I have approached it for years because I am uneducated when it comes to understanding phase. And I have felt like I have been missing something in my DIY home speaker designs. I am attempting to address this by:

1. No more 4-way designs. Instead, one big planar for 200hz-4khz to address human vocal range and most instruments. I do not like large full range drivers due to a number of issues so that is not an option. And I love ribbons so they will round out the top end. TA will come into play for blending with woofers at 200hz but not a big deal being under 300hz. Will need to play with the higher xover point to match dispersion...

2. 1st order x-overs...easy for the 4khz xover, not for the 200hz xover...might bump to 300hz and try for 1st order...or rely on RePhase.

3. Rephase. I would also like to try Dirac though that is $$$. But I will attempt the trial.

I am also considering bagging my grandiose ideas and just get big Magnepans like 3.7 with the dedicated ribbons. Mmmmmmm. Just maybe...
 

·
Registered
Joined
·
5,098 Posts
You heard and liked the Magic Bus right? That is all 4th order LR. Sure there are a lot of other things going on in the bus and even the environment it is very different to our daily drives. So fourth order can be made to sound good, even in the region where we are most sensitive to phase. You highlighted the quality of low end in the bus. Again, the size of the bus has a lot to do with how accurate and articulate you can get the low end. But yeah, it's a fourth order LR. Timing and response coherence are key, phase is accounted for between those two. Anything beyond that is a mirage.

Our hearing is geared to hearing relative differences in timing, amplitude, and pitch rather than absolutes. Anytime we saturate one area eg listening at 120db (amplitude) the ears ability to hear 'difference' degrades rapidly, which is why we won't hear 10% distortion at 120db etc.

By running shallow slopes in a car, not only are you spreading the combing but also making a comb with much finer teeth. So you are actually saturating two fields pitch and amplitude (peak/dip). You're pushing the ear out of its comfort zone, chasing something that is a mirage in this environment.
 
1 - 20 of 115 Posts
Top