ProSoundWeb Community

Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => Audio Measurement and Testing => Topic started by: Alberto Escrina on November 08, 2014, 01:39:32 PM

Title: Crossover phase handling
Post by: Alberto Escrina on November 08, 2014, 01:39:32 PM
I know, this topic must be touched a hundred times. I was searching in the archives for a couple of days and didnīt find anything that matches.
We all know that, a basic two way system with sources with different arrival times for example, can be time aligned keeping the low frequency driver, which in this example has an earlier arrival time (it is closer to the listener), whit no signal delay added and inserting "extra" electronic delay to the high frequency driver, even he has an already later arrival time.
We match phase on both at the acoustical crossover point and thats it. A flat crossover transition.
The result of this is that we have an "X" style of phase traces from the two of them matching at crossover frequency but moving away very quickly as frequency moves up and down.
On the other hand we can do the obvious, to delay the closer low frequency woofer and, of course, we will have an extended portion of the phase trace with similar slope matching nicely.
Adding both signals give us a confirmation that, at the zone of interest, the later seems to give better results.
As one of the old schoold guys from the days we didnīt have any way of handling phase except with passive or active all-pass filters, I tend to think that is not a good decision to add electronic delay to the low frequency section of our systems. Why? I dont know. But, believe me, it was something "everybody" knows.
After the DSP revolution, we keep the woofers with no delay added and insert it to the other sections on an 360° extra shift basis.
Is just an old man manía or am I missing something I once knew?
And, yes, I work for a manufacturer.
Regards from Argentina!
Title: Re: Crossover phase handling
Post by: John Roberts {JR} on November 08, 2014, 01:51:19 PM
You are thinking of this as a two dimensional problem but loudspeakers combine in a 3 dimensional space so perfect time/phase alignment on one axis will not be perfectly aligned for other axis.

This is an old and well explored subject, look up crossover design. While not trivial, the basics are well explored.


JR
 
PS: Bienvenido,  I just noticed today that my frozen blueberries came from Argentia so gracias.
Title: Re: Crossover phase handling
Post by: Alberto Escrina on November 08, 2014, 02:21:23 PM
You are thinking of this as a two dimensional problem but loudspeakers combine in a 3 dimensional space so perfect time/phase alignment on one axis will not be perfectly aligned for other axis.

This is an old and well explored subject, look up crossover design. While not trivial, the basics are well explored.


JR
 
PS: Bienvenido,  I just noticed today that my frozen blueberries came from Argentia so gracias.

Thank you JR!
Iīm glad you can taste our blueberries! De nada!
Yes, Iīm aware on the implications of phase on on and off-axis reproduction.
We do have that kind of research for our products. Just one example for our website:

http://www.stsproaudio.com/pdf/MedicionesConcertoTop.pdf

I also understand why that "X" shape of phase traces is produced in one case on not in the other, as well as other implicances on this issue.
What I wonder is: do designers apply digital delay to the low frequency section if needed or thereīs an implicit convention or some other reason for not doing so?
Thatīs all.
Thanks again.
Regards from Argentina!


Title: Re: Crossover phase handling
Post by: John Roberts {JR} on November 08, 2014, 02:30:18 PM
I don't do loudspeaker design (but did some crossover design a few decades ago). As I understand it, for drivers mounted on a flat baffle board the effective source of the LF drivers would be inches behind the plane of the baffle board so further away from the listener than the HF drivers. If anything the HF drivers would need delay to time align them. But we are talking about small amounts of delay and/or phase error at long wavelength LF.

Sorry I am not a speaker expert so maybe somebody else can be more helpful. 

JR
Title: Re: Crossover phase handling
Post by: Mac Kerr on November 08, 2014, 02:30:59 PM
Thank you JR!
Iīm glad you can taste our blueberries! De nada!
Yes, Iīm aware on the implications of phase on on and off-axis reproduction.
We do have that kind of research for our products. Just one example for our website:

http://www.stsproaudio.com/pdf/MedicionesConcertoTop.pdf

I also understand why that "X" shape of phase traces is produced in one case on not in the other, as well as other implicances on this issue.
What I wonder is: do designers apply digital delay to the low frequency section if needed or thereīs an implicit convention or some other reason for not doing so?
Thatīs all.
Thanks again.
Regards from Argentina!

I think the main reason for not adding delay to the low frequency side of a crossover is the inherent "delay" caused by the phase rotation due to the low pass filter. If you are referring to a pure impulse response the output of the low pass will appear to be delayed by the amount of phase change at the frequency of interest. At 100Hz 360š of phase offset is one wavelength, which is about 10', which is about 10ms. A smaller amount of phase offset will have smaller latency, but those are pretty significant delay times when you are talking about time aligning the components of a mulitway speaker system.

Mac
Title: Re: Crossover phase handling
Post by: Ivan Beaver on November 08, 2014, 02:32:29 PM
s.
We all know that, a basic two way system with sources with different arrival times for example, can be time aligned keeping the low frequency driver, which in this example has an earlier arrival time (it is closer to the listener), whit no signal delay added and inserting "extra" electronic delay to the high frequency driver, even he has an already later arrival time.
We match phase on both at the acoustical crossover point and thats it. A flat crossover transition.

Why do you "assume" the woofer will "arrive" first?

Yes it may be physically closer-but there are elements that often make it "arrive" later.

The low pass section of crossover adds delay/phase shift.  And the large inductance of the woofer voice coil (as compared to the much smaller inductance of the HF driver) adds additional "delay" phase shift.

The actual crossover freq used will also affect how much delay difference there is.

Lets say you have 2 drivers (it could be a woofer and a tweeter or a sub and a full range cabinet).

Lets assume you have a "perfect" alignment-which is flat phase through the crossover region.

Now change the crossover freq.  The amount of delay needed to align the devices will change.

Actually you need to have the phase aligned for a good region ON EITHER SIDE of the crossover freq.  I look for about 1 octave on either side.

The reason is that the sound does not "stop" once it goes above or below crossover.  It is a SLOPE, and the different freq bands WILL interact with each other until the difference is large enough in level that they don't.

It is real common to get a "perfect" phase match at crossover-yet have a notch in the response above or below crossover freq because the phase is not lined up as well there.

My "quick test" to see if the alignment is done well is to save a trace of each of the freq bands after alignment.  Then  turn them both on and see if you get addition across a 2 octave range.

If you do then GOOD.  If not, then you need to look closer at the alignment-not only in time-but in filter types/slopes/freq/eq etc.

Now since you are getting good addition, you need to also deal with that to make sure the amplitude is flat-assuming that is the design goal.

Sometimes all of this is real easy and quick (once you get the hang of it) and other times you can spend A LOT of time trying to get a good alignment.

ALSO remember when you are dealing with a cabinet that has a woofer and a tweeter that are offset (as most devices are), the alignment you come up with is ONLY good for that listening position.  If you change listening positions, you will need a different alignment.

So you have to figure out some sort of "average" or target where you want it aligned to, and be willing to accept that all other positions will be "wrong".

The only time a single alignment will be good for all listeners (within the intended coverage pattern) is when the distance from the 2 freq bands is the same to all the listeners.  Even small parts of a ms matter-maybe.  It depends on the freq of interest.

As you go up in freq the differences get VERY small.  Down low (say between a sub and woofer) you can move around a good bit and not require a different delay freq.

It all depends on the wavelength.  Higher ones are harder to deal with-because the wavelength is so short.
Title: Re: Crossover phase handling
Post by: Alberto Escrina on November 08, 2014, 02:57:12 PM
Hi all and thank you for your answers.
Ivan, I donīt assume, I measure it. Looking to both, impulse responses and their phase difference at the crossover point.
As you and Mac correctly add, those phase values are because of physical position, filter and crossover phase shifts, etc.
I agree with everything you say.
My aligning procedure is similar to yours. No disagree with that.
This is not the grey area of my concern.
My only question is:
Do designers apply digital delay to the low frequency section of a multi way system or thereīs an implicit convention or some other reason for not doing so and insert whatever it works on the other sections?
Thank you guys.
Regards from Argentina!
Title: Re: Crossover phase handling
Post by: Mac Kerr on November 08, 2014, 03:06:44 PM
My only question is:
Do designers apply digital delay to the low frequency section of a multi way system or thereīs an implicit convention or some other reason for not doing so and insert whatever it works on the other sections?

I think both Ivan and I pointed out that the low frequency already has latency caused by the low pass filter, which is why the upper band needs to have delay added to get the 2 signals "in phase" at the acoustic crossover. Adding electronic signal delay to the low passed band will only make the problem worse.

Mac
Title: Re: Crossover phase handling
Post by: Alberto Escrina on November 08, 2014, 04:57:44 PM
I think both Ivan and I pointed out that the low frequency already has latency caused by the low pass filter, which is why the upper band needs to have delay added to get the 2 signals "in phase" at the acoustic crossover. Adding electronic signal delay to the low passed band will only make the problem worse.

Mac

Yes you did.
But yours is an assumption and Iīm providing facts.
The 1.4" compression driver is mounted in a 20cm long wave shaping device. The L.F. driver is a 10" cone.
Impulse measurements confirm, with HPF and crossover engaged, that the cone driver arrives first by far. Iīm at home and donīt remember figures.
But forget for a moment this specific example and suppose that you do have a situation like this, would you be so kind to read my first post where I describe two alternatives to put them "in phase" and then again my main question is:
Do designers apply digital delay to the low frequency section of a multi way system or thereīs an implicit convention or some other reason for not doing so and insert whatever it works on the other sections?
Thanks a lot guys!
Regards from Argentina!
Title: Re: Crossover phase handling
Post by: Mac Kerr on November 08, 2014, 05:39:53 PM
But forget for a moment this specific example and suppose that you do have a situation like this, would you be so kind to read my first post where I describe two alternatives to put them "in phase" and then again my main question is:
Do designers apply digital delay to the low frequency section of a multi way system or thereīs an implicit convention or some other reason for not doing so and insert whatever it works on the other sections?

Maybe you should reread my post. It is not an assumption, it is a fact. The low pass filter causes an apparent time offset in the low frequency signal which puts it significantly behind all other drivers in the system so it would be self defeating to add delay to the low frequency band of a multiway system. It is not an implicit convention, it is common sense. You correct each band pass back to the lower frequency of the pair. Unless there is a large physical offset between the drivers relative to the frequency the lower frequency driver will always be behind the higher frequency driver.

There are 2 schools of though on how to "time align" pairs of drivers, matching phase through the acoustic crossover region, and impulse response. Since the wavelength of the low frequency low pass filter of a woofer or subwoofer could be from a couple of feet to 10-15 feet it is hard to measure the impulse response, so since the advent of dual FFT measurement systems that make it simple measure phase response it is more common to align speaker systems for good phase response through the crossover region. In either case, the acoustically measured time offset will always have the lower frequency driver trailing, and the higher frequency driver requiring delay to get back in phase.

Mac
Title: Re: Crossover phase handling
Post by: Ivan Beaver on November 08, 2014, 08:48:15 PM
and then again my main question is:
Do designers apply digital delay to the low frequency section of a multi way system or thereīs an implicit convention or some other reason for not doing so and insert whatever it works on the other sections?
Thanks a lot guys!
Regards from Argentina!
So the ANSWER to that question is NO, because it is the HF driver that needs to be delayed-for the reasons Mac and I gave.

Impulse alignment is NOT the way to get a good transition between devices.  PHASE alignment is the proper way to get the best sounding alignment.
Title: Re: Crossover phase handling
Post by: Peter Morris on November 08, 2014, 10:29:38 PM
So the ANSWER to that question is NO, because it is the HF driver that needs to be delayed-for the reasons Mac and I gave.

Impulse alignment is NOT the way to get a good transition between devices.  PHASE alignment is the proper way to get the best sounding alignment.

Not always, here is an example http://eaw.com/docs/2_Legacy_Products/Processor%20Settings/KF750/KF700_Series_PROCS_rev1.pdf

Some times the LF needs to be delayed.
Title: Re: Crossover phase handling
Post by: Merlijn van Veen on November 09, 2014, 03:49:15 AM
Hi Everybody,

Maybe the slides from my seminar will help in illustrating the pitfalls from aligning a crossover with impulse response instead of phase.

As has been pointed out, the introduction of filters introduce phase shift on top of physical displacement. This also affects the appearance of the impulse response. Especially for low-pass filters. The impulse response gets stretched over time and amplitude "appears" to have been reduced.

When you align the impulse peaks of both drivers you basically align the high frequencies of the HF driver to the low frequencies of LF driver. As can be seen in the phase (group) delay slide. Chances are you'll achieve the opposite of your goal, summation at XO.

When you align with phase. The "low" frequencies of the HF driver will connect to the "high" frequencies of the LF driver, resulting in summation at XO.

disclaimer: In passive designs where delay is not available. Either driver might lead by physical displacement and/or design. Summation can still be achieved by a combination of 0.5, 1.5, etc... wavelengths out of polarity or 1, 2, etc... wavelengths in polarity phase shift. The payoff is response ripple throughout the XO region proportional to the difference in cycles.


Regards,


Merlijn
Title: Re: Crossover phase handling
Post by: Ivan Beaver on November 09, 2014, 09:18:03 AM
Yes you did.
But yours is an assumption and Iīm providing facts.
The 1.4" compression driver is mounted in a 20cm long wave shaping device. The L.F. driver is a 10" cone.
Impulse measurements confirm, with HPF and crossover engaged, that the cone driver arrives first by far. Iīm at home and donīt remember figures.
But forget for a moment this specific example and suppose that you do have a situation like this, would you be so kind to read my first post where I describe two alternatives to put them "in phase" and then again my main question is:
Do designers apply digital delay to the low frequency section of a multi way system or thereīs an implicit convention or some other reason for not doing so and insert whatever it works on the other sections?
Thanks a lot guys!
Regards from Argentina!
Let's look at it from a little math approach.

Let's assume a 1Khz crossover and use assume 1Kz is 1' long (just to make it quick and easy). 

Each pole of a crossover is a 45° shift or 0.125' (at 1Khz)

Let's also use 24dB filters (4 poles).  The low freq filter will "move" the woofer back so it moves back 0.5'. 

The high freq filter will "move" the tweeter forward 0.5'.  So now (in most cases) the woofer is "physically behind the tweeter. 

And let's don't forget the possible additional 0.125' lag to the woofer because of the self inductance of the voice coil itself.

So the Tweeter is now "in front" (at least as far as phase is concerned) so it needs to be delayed-NOT the woofer.

Yes the actual physical layout of the cabinet will change things-but it is not a simple physical position that makes you delay one driver vs another.  It is the phase of the drivers at a particular position.

So the real answer is "It depends"
Title: Re: Crossover phase handling
Post by: Ivan Beaver on November 09, 2014, 09:20:21 AM
Not always, here is an example http://eaw.com/docs/2_Legacy_Products/Processor%20Settings/KF750/KF700_Series_PROCS_rev1.pdf

Some times the LF needs to be delayed.
Agreed-it depends on the particular layout of the particular cabinet.

But "in general" it is the HF that needs to be delayed.
Title: Re: Crossover phase handling
Post by: Alberto Escrina on November 16, 2014, 03:47:22 PM
Hi everyone,
Finally I have some time to thank you all for your contributions.
Unfortunately the request I presented in my first post took a different turn than intended.
I never actually asked if, in my example, it was the low frequency driver the one that must be delayed or not. I just presented an example when it is.
I never actually said that impulse align was better than phase align or viceversa.
Since English is not my native languague, as you might have noticed already, I might have not expresed myself properly. Sorry about that!
This is why I was asking you to reread my first post. Because the core of the issue was another one.
Anyway, Peter has provided one example that answers my question. EAW does.
Thank you very much for posting and sorry for those misunderstandings!
Best regards,
Title: Re: Crossover phase handling
Post by: Doug Fowler on November 17, 2014, 04:52:45 PM

The low pass filter causes an apparent time offset in the low frequency signal which puts it significantly behind all other drivers in the system

Mac

This is widely misunderstood.  I do a demonstration in my classes which easily demonstrates this.  Open the LPF to over 1 KHz, set your analyzer delay locator on "auto" and observe the latency.  Then, begin to lower the LPF down to say, 100 Hz and watch the latency increase.  It's also interesting to open an IR window and watch the IR begin to spread out and lose apparent amplitude due to the apparent "time stretching" of lower frequencies.
Title: Re: Crossover phase handling
Post by: Robert Lofgren on November 17, 2014, 05:34:43 PM
Anyone who has any insight about the Le Cleach crossover theory and if it would be suitable for live sound?

http://www.audioheritage.org/vbulletin/showthread.php?4850-A-New-Crossover-Theory