acoustic phase question

Status
Not open for further replies.
Jump to Latest
C

crazyhub

Member
Joined 2005
#1
Hi,

Let us suppose a classical two-ways.
Three different filterings (slopes and\or Fx) show, under identical conditions of measure, a same or very close magnitude response.
The first filter shows a crossing of the acoustic phases of each driver exactly at Fx. Both drivers are thus exactly " in phase " at Fx.
The second filter shows no crossing of the phases at Fx, but close and parallel phases on a wide band of frequencies on both sides of Fx.
Third filter shows no crossing phases but phases going away one of the other one.


What do these differences of phases between drivers mean in physical terms, as well as their corollary in the listening?

Thank you.😉
 
#3
Syncroniq said:
Hi,

Instead of looking at the phase, you can invert the polarity of the tweeter. If the summed response have a huge suckout at XO, the drivers are well in phase.
Click to expand...
Hi,
Yes I knew this reverse nul effect; my question was about the effects of drivers not perfectly in phase at Fx however giving a straight fr magnitude.

:scratch:
 
#5
Hey,

You can have a response from the mid and tweeter, giving a flat response, but not being in phase. Thats how alot of companies does it, but it doesnt sound good. image and perspective depends on phase.

The best way is the reversed null method. flat in phase, and deep null out of phase, and you have a very good startingpoint.
 
#6
Syncroniq said:
Hey,

You can have a response from the mid and tweeter, giving a flat response, but not being in phase. Thats how alot of companies does it, but it doesnt sound good. image and perspective depends on phase.

The best way is the reversed null method. flat in phase, and deep null out of phase, and you have a very good startingpoint.
Click to expand...

However, that is only true for even order XO's. To sum properly, odd order XOs, i.e. 3rd order butterworth, the high pass and low pass have to be 90 degrees apart which doesn't yeild a null when reversed.

Regards,

Dennis
 
#7
sreten said:
Hi,

It is a none question, different filters cannot sum the same.
Define the filter functions with the properties you describe.
2 and 3 seem not feasible.

🙂/sreten.
Click to expand...
I spoke of different filters using different slopes and/or Fx suming the same fr magnitude, no matter the functions; thus 2 and 3 are feasible, no?
 
#8
djarchow said:


However, that is only true for even order XO's. To sum properly, odd order XOs, i.e. 3rd order butterworth, the high pass and low pass have to be 90 degrees apart which doesn't yeild a null when reversed.

Regards,

Dennis
Click to expand...
That's interesting; so can I consider that if the summed phase looks unshifted + wanted (flat) fr magnitude, it's OK, at least for this type of filter?
 
#9
Hi,

In praksis, thats not quite true.

You will have a timedelay between the two drivers, so you will never be put in that situation as there will be a timedelay between the two drivers. If you get a flat response in phanse, and not "nulling" out when out of phase, the filter will not perform optimal. The drivers NEED to bee in phase at least at xo point. It will be reflected in stereoperspective, staging and so on.

Just becourse its "a type of filter", doesnt mean its good. There are thousands of "types of filters", most of them just havent been named yet. 🙄

You should forget about butterworth, bessel and so on. Its way to theoretical to work in pracsis. Find the rolloff that makes your drivers sum flat in phase, and null out when out of phase. You might end up with slopes of 27dB/oct on the tweeter and 14dB/oct on the bass (just an example).
 
#10
Hi,

Here you can see the slopes i came to, in order to get flat response, and good phase, on a 3 way.

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.


If your drivers are timealigned, and have a flat response to begin with, Steen duelund have developed a filter that are in phase at all frequencies, and give a flat response. Read more here.

http://www.steenduelund.dk/download/duelund-filter.pdf
On his site, theres other interesting articles. He past away a couple of years ago.
 
#12
Syncroniq said:
Hi,

In praksis, thats not quite true.

You will have a timedelay between the two drivers, so you will never be put in that situation as there will be a timedelay between the two drivers. If you get a flat response in phanse, and not "nulling" out when out of phase, the filter will not perform optimal. The drivers NEED to bee in phase at least at xo point. It will be reflected in stereoperspective, staging and so on.
Click to expand...

This is correct for even order acoustic slopes. The more correct answer is that the acoustic response of the drivers and XO have to have the proper phase relationship between them to sum properly based on their slopes. That may be in phase as you mentioned and certainly will if you have even order slopes, but if your acoustic slopes are odd order then they need to be 90 degrees apart to sum properly.

Yes interdriver delays are a common problem in any design, but there is nothing about those delays that says that your final slopes have to be 2nd order, 4th order etc.

Syncroniq said:

You should forget about butterworth, bessel and so on. Its way to theoretical to work in pracsis. Find the rolloff that makes your drivers sum flat in phase, and null out when out of phase. You might end up with slopes of 27dB/oct on the tweeter and 14dB/oct on the bass (just an example).
Click to expand...

Mixed order high pass and low pass are a common way to deal with interdriver delays. The exact same approach works just as well for odd order slopes as even order. Odd order slopes (which should never be in phase at fc) can image and perform just as well as even order slopes; just ask Jim Thiel and Joe D'Appolito. And before anyone jumps on me, yes I know that Joe D has moved to 4th order LR slopes for mtm's these days.

You are correct that there are thousands of filter types out there, but by limiting yourself to only those that are in phase at fc you are missing out on half of them.

Kind regards,

Dennis
 
#13
OK guys I do understand the effects of phase on the distributed energy out of axis and its influence in room; what I don't understand is why filter one is commonly regarded as advantageous vs filter two:
<<The first filter shows a crossing of the acoustic phases of each driver exactly at Fx. Both drivers are thus exactly " in phase " at Fx.
The second filter shows no crossing of the phases at Fx, but close and parallel phases on a wide band of frequencies on both sides of Fx.>>
:scratch:
EDIT : quote from Dennis: " Odd order slopes (which should never be in phase at fc) "...Seems very logical to my mind for what I understood of this matter! indeed trying to make the phases closer on a given bandwidth, no?
 
#14
Again it depends on how the high pass and low pass are supposed to sum.

For even order XOs the high pass and low pass phase should be "in phase" over as wide a range as possible on either side of Fc.

For odd order XO slopes the high pass and low pass phase should be 90 degrees apart over as wide a range as possible on either side of Fc.

A goal for any XO is that the acoustic phase of each driver and XO matches the target phase over as wide a range as possible. Now as Syncroniq points out, driver acoustic offsets and their associated delay may force the designer to steepen or relax one of the drivers slopes to get good summation and phase alignment

In response to your question about two drivers phase that is close and parallel over a wide range, may imply that phase alignment is probably a bit better above or below the the measurement or design axis.

Regards,

Dennis
 
#15
Hi,

GM> Great article. But if "signal alignment" is not possible, you have to vary the slopes, instead in order to get a good phase.

The articles result is not perfect, as the "nulling" isnt symetrical. And the "zero time point" (or what its called in english) is about (but not exactly) where the voicecoil former is connected to the cone.

If time alignment where an option, i would use the duelund alignment, as it brings the drivers in phase over a wide areal. Where as the article example is only in phase just about at xop. Just belov, there not in phase (if i read the article right).

I would recommend downloading lspcad demo and then simple play with it (lspcad) until you see how filterorder and phase affects/interact.

http://www.ijdata.com/LspCAD_demo.html

You can download my meassurements here, to import and play with filters directly in lspcad.

2way scanspeak 15W/R29:
http://www.speakerbuilder.dk/doc/articles/Passage10Mess.zip

Heres my filter for the above speaker:
An externally hosted image should be here but it was not working when we last tested it.


there are more on my website.

Try and make a 1. order, then 2. order and so on. Then mixed order (3. order on tweeter and 2. order on bass) and see how the phase is doing, when the rolloffs fits.

Its free 🙂
 
#16
OK guys, thank you; let's go a little further please:

Dennis you wrote <<<A goal for any XO is that the acoustic phase of each driver and XO matches the target phase over as wide a range as possible.
***what do you mean by "the target phase"?

Syncroniq: thank you for your lspcad proposal but I already work with SW which am used in and pleased with.

Here are shoots of my last 3-way filter chart and filter diagram simulations; I decided to show you the relatives phases only to simplify the reading; it's hard for me to be sure about the slopes of the tweeter HP and mid LP but seems to be even order types (24/12db) and as you can see the phases are very close on a wide bandwidth and even crossing close to fc.
However the phases between the mid and the woofer aren't close (+140° / -140° thus about 80° difference); baffle is vertical and not slanted so the drivers acoutic centers aren't aligned. However drivers are close spaced to minimise directivity problems. I have to say that I'm very pleased with the listening result with this filter and will not change it: notably soundstages and imaging are, at least to my opinion, of first class.
I just wasn't intellectually satisfied because of these relative woofer/mid phases distances; regarding that the woofer LP and mid HP are mixed order, this distance seems not to be a real problem, isn't it?
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
 
#17
crazyhub said:
OK guys, thank you; let's go a little further please:

Dennis you wrote <<<A goal for any XO is that the acoustic phase of each driver and XO matches the target phase over as wide a range as possible.
***what do you mean by "the target phase"?
Click to expand...


When you design a XO you would normally decide upon a target XO point and slope, lets say 2500 Hz and 4th order Linkwitz Riley. I think in SpeakerWorkshop they call this a "Goal". Then you design your XO to match your acoustic slopes to those target slopes as closely as possible.

For example, take the 4th order Linkwitz Riley; along with the specific spl rolloff the phase relationships are defined as well for the woofer and tweeter to be in phase at all frequencies. Of course outside of theory, this isn't really possible beacuse at some point your drivers will no longer be able to match your target rolloffs which also means your phase won't match either. The same is true of inter driver delays. Generally though, the best performing speakers maintain the proper phase relationship between the drivers over as wide a range as is possible.

Hope this helps

Regards,

Dennis
 
#18
djarchow said:



When you design a XO you would normally decide upon a target XO point and slope, lets say 2500 Hz and 4th order Linkwitz Riley. I think in SpeakerWorkshop they call this a "Goal". Then you design your XO to match your acoustic slopes to those target slopes as closely as possible.

For example, take the 4th order Linkwitz Riley; along with the specific spl rolloff the phase relationships are defined as well for the woofer and tweeter to be in phase at all frequencies. Of course outside of theory, this isn't really possible beacuse at some point your drivers will no longer be able to match your target rolloffs which also means your phase won't match either. The same is true of inter driver delays. Generally though, the best performing speakers maintain the proper phase relationship between the drivers over as wide a range as is possible.

Hope this helps

Regards,

Dennis
Click to expand...
Yes Dennis, you're of great help!
Just to be sure because of my poor english, when you're talking about even order acoustic slopes, these are 2nd (12db) or 4th (24db) or 6th (36db)?
And odd order are 3rd (18db) or 5th (30db) and on?

If I decide to go with mixed orders (ie 2.5way or full-band+tweeter with the simpliest filter possible), by definition I will not have to look for crossing phases, no?
 
Status
Not open for further replies.
Top Bottom