Active Crossovers - A different perspective

[DrWho]

I think it was Tom (or maybe Shawn) who mentioned a while back that time-alignment also affects polar lobing. This is something I hadn't thought of before, but a little light went on in my head as to the possible implications. After a little research, I came across this page from the Rane website, which is discussing active linkwitz-riley crossovers:

http://www.rane.com/note160.html

Lo and behold, they provide all sorts of cool information regarding the effects of steering your polars. For those that just like pictures, I thought I'd sum up the article and post the pictures here.

Here we have the polar response of a time-aligned system using a butterworth filter:

(notice that the main center lobe is tipped down)

And then here we have the response of a time-aligned system with a Linkwitz-Riley crossover:

(Notice that the main center lobe is firing straight)

And then here we have a linkwitz riley with and without time-alignment:

After reading this article I've determined that I need to go do some more listening, but I have to wonder if the most significant audible aspect to time-alignment isn't just the shift in polar response. I also wonder if this doesn't possibly explain why some people don't hear as much of a difference...changing the polar response changes how the speaker loads the room, which may or may not introduce significant changes in the sound (but at that point, it might be argued that a system that reveals more significant changes is probably a better acoustical environment).

Thoughts anyone?

[Guest " "]

that makes sense...you won't hear a difference unless you are standing in the lobe.

[DrWho]

you won't hear a difference unless you are standing in the lobe.

The point is that the lobe is moving...you can't just stand in it. You have to move with it.

[PrestonTom]

"....

After reading this article I've determined that I need to go do some more listening, but I have to wonder if the most significant audible aspect to time-alignment isn't just the shift in polar response. I also wonder if this doesn't possibly explain why some people don't hear as much of a difference...changing the polar response changes how the speaker loads the room, which may or may not introduce significant changes in the sound (but at that point, it might be argued that a system that reveals more significant changes is probably a better acoustical environment).

Thoughts anyone?

...."

Interesting plots Mike.

Regarding the issue of whether it the most audible aspect of time alignment, that might be a difficult one to tease out. There was a study in JAES a couple of years ago showing that folks are sensitive to group delay in the spectrum per se. That is when there were no overlapping bands with different phase spectra. There has not been a great deal of research, but what has been shown is that folks can be sensitive to group delays (offsets) of a few msecs or greater. The sensitivity (amount of group delay required for detectability) was dependent on where in the frequency spectrum the group delay occurred. Interestingly, the listeners' sensitivity was decreased if the play back was in a reverberant environment (less sensitivity to the group delay).

Good Luck,

-Tom

[DrWho]

Have you ever seen mention of offsets less than 1ms introducing large audible differences?

[PrestonTom]

Have you ever seen mention of offsets less than 1ms introducing large audible differences?

An offset where there would be very little or no overlap between the adjacent bands? Hmmm? There was once a conference paper that I heard but the topic was not a straight analogy to the issue of group delay. So I won't go into it.

Of course, if the bands (driver outputs) do overlap spectrally you would be getting also sorts of cues (delay & add leading to comb filtering etc). This is why steep filters are so attractive.

-Tom

[DrWho]

The bands were overlapping...basically I was toying with time-alignment on my Chorus II's, which have rather short time offsets between the squawker and the woofer/tweeter. Both my lab partners and I would describe it as sounding less congested and more clean, which I would expect to be the result of less time smear through the passbands - we were listening to some music that hit right in the crossover frequency, which I think was the most important aspect.

We were testing with 4th order Linkwitz-Riley (24dB/octave), but I can do 8th order (48dB/octave) now so when I have time I'm going to try to compare the relative magnitude of change. If I have more time I intend to implement 16th order too (96dB/octave), but I'm kinda worried about the group delay trade off....so I might try experimenting with some non-causal filters to remove the group delay altogether (something a Shure engineer told me was possible, although I have no clue how to do it). Oh the beauty of DSP.

[Guest IVstringer]

If you look at it from a pure "time shift causes a phase shift" perspective, try out this scenario:

f = 3kHz ==> T = 0.33 ms

So, a time shift of only 0.15 ms means you shift about 180 deg. You could go from in phase to out of phase with only that much adjustment. Of course, things are more complicated than that, but it would suggest that it's entirely possible to have this behavior occur. Granted, you can't be so picky because you go to a different frequency and the phase alignment is all different already.

[mas]

If you look at it from a pure "time shift causes a phase shift" perspective, try out this scenario:

f = 3kHz ==> T = 0.33 ms

So, a time shift of only 0.15 ms means you shift about 180 deg. You could go from in phase to out of phase with only that much adjustment. Of course, things are more complicated than that, but it would suggest that it's entirely possible to have this behavior occur. Granted, you can't be so picky because you go to a different frequency and the phase alignment is all different already.

Be very careful with this analogy. Phase response of a single source and superposition are two very separate issues. And phase shift in a single source means little until you reach excessive rates of change (the slope or dx /dt of the slope) in the phase (group delay).

The polar anomalies will result from multiple sources, be they real or virtual (diffractive sources or reflections), source the same frequency signals (the same passband).

BTW, what Doc mentions is exactly the reason behind the original D'Appolito MTM configuration - whereby the odd order LR filter resulted in an ~15 degree upward polar tilt. The addition of the additional M unit on top tends to nullify this lobing and through cancellation tends to tilt the polar lobe back on axis. I have not personally measured this to see what the true result is, as seldom to you add more superposition to actually result in a more ideal situation rather than simply tending to shift the major charactristics in different direction, albeit with allot of undesired spurious artifacts.

And yes Doc, just increasing the slope of a crossover is not the panacea! Phase wrap can cause lots of 'funny' problems! This was quickly discovered when agile active crossovers became available. And this is readily apparent in large format sound systems where you have ample space to ramble about and experience the polar dispersion in the far field.

Signal alignment indeed effects polar lobing in 3 space! Hence the advantage of a vertical stacking of drivers versus a horizontal alignment. You are choosing the degrees of freedom of the system where one axis remains 'contant' (relatively speaking) and the other axis is adjustable. The vertical stacking of drivers convention serves to keep the generally well behaved vertical lobing configuration 'constant' while the more chaotic horizontal polar lobing is adjusted.

The only way to completely address this is via totally coincident (ie: coaxial) drivers - be they real or virtual. Thus we deal with the practical limitations of not having a completely satisfactory coincident physical source. Short of this, we chose the arrangement of drivers and then apply signal alignment in the time domain to reduce the affects of superposition in the system.

In a system with significant acoustic origin offsets, signal alignment is a 'necessary' addition to the system...something a passive crossover cannot provide. Hence another reason that an active crossover with signal alignment should be considered fundamental to a Heritage series speaker.

BTW2: The Rane technical references are an excellent source of info!

[mas]

Have you ever seen mention of offsets less than 1ms introducing large audible differences?

An offset where there would be very little or no overlap between the adjacent bands? Hmmm? There was once a conference paper that I heard but the topic was not a straight analogy to the issue of group delay. So I won't go into it.

Of course, if the bands (driver outputs) do overlap spectrally you would be getting also sorts of cues (delay & add leading to comb filtering etc). This is why steep filters are so attractive.

-Tom

This may help a bit. There are three

[mas]

The second.

[mas]

The third

[Coytee]

Go MAS go!!!

[DrWho]

If you look at it from a pure "time shift causes a phase shift" perspective, try out this scenario:

f = 3kHz ==> T = 0.33 ms

So, a time shift of only 0.15 ms means you shift about 180 deg. You could go from in phase to out of phase with only that much adjustment. Of course, things are more complicated than that, but it would suggest that it's entirely possible to have this behavior occur. Granted, you can't be so picky because you go to a different frequency and the phase alignment is all different already.

Going one step further, every integer multiple of .33ms will result in a flat frequency response...

My measurement rig isn't able to provide absolute time delays, only relative within the same measurement (because it automatically shifts time = 0 to the largest peak in the impulse). This is annoying because it prevents one from measuring the squawker, tweeter, and woofer independantly to figure out the time offsets. I pulled out a ruler and figured I'd start with the theoretical time offset as a function of distance and then walk the delay down until I achieved +6dB summation at the crossover frequency. I forget the actual numbers right now, but we basically ended up delaying the tweeter not enough by one period...or at least that's what we think because a certain snare hit sounded very blurred. We shifted by something like 200 microseconds and the snare sounded clean. Going one more period further, the snare sounded blurred again, so we decided the cleaner one was probably the correct time offset. I wanna say we came to something like 800 microseconds for the delay?

[mas]

that makes sense...you won't hear a difference unless you are standing in the lobe.

Actually you wont hear a difference unless you are NOT in the lobe.

Remember, the lobe is not constant, the Q of each lobe, and the number of lobes are frequency dependent - in other words, the lower the frequency, the lower the Q (wider the lobe) and the fewer there are, and the higher in frequency the higher (narrower) each lobe's Q and the greater the number of lobes.

Hence, if you are 'in' a lobe, you hear the signal. If you are in a null, where the out of phase signals cancel, you will not hear that component of the total signal, and as Doc says, the lobes vary with frequency. And they will be different in each part of the room.

The irony, is that this characterisitic is what Bose has traditionally used to their advantage! Their direct -reflecting signal speakers resulted in a chaotic soundfield, where all areas were universally mediocre. Unlike more traditional speakers where you could more clearly identify an on axis sweet spot and an off axis null, with Bose, almost every spot was some complex combination of lobes and nulls. Thus, as you moved from one spot to another you did not experience a sweet spot and a 'bad' spot. Hence the uniforminty (without reference to a trully good sweet spot) seemed to result in a 'wider' sweet spot, as you didn't notice the contrast! And without comparison to what it SHOULD sound like, its all good. After all, bad is relative to what is good. And without a good, you don't have a bad - at least apparently!!!

An easy place to experience this, and my guess why many are not readily aware of the affects of polar anomalies, is that in a small room, one does not often become aware of the nulls as they are close to the speakers (but the effect is still present! - but like the Bose, just not as distinct). In a live sound arena prior to a show, it is easy to walk about the room in the far field, and the nulls become large and VERY Readily apparent!

The irony is further aggravated by the traditional misuse of EQ (which was traditionally viewed as a solution), whereby the problems were NEVER resolved! Instead what was felt to be a solution was simply the addition of additional phase manipulation (advance or lag) to the direct signal due to the use of the ®LC filters employed in the EQ filters!

The result was that the change modified the signals such that the resultant superpostion simply caused the lobes to shift slightly to the right or left, until the null was moved off of the listening position (and thus apparently solving the problem) and onto the poor suckers who paid good money for their seats, but who for all intents were clueless as to why the sound simply sounded "live", and that is what many associate as a fundamental component of what makes up the "live" sound! And Gee! Wasn't that concert great!!!(?)

[boom3]

"universally mediocre"--love that phrase! As in, crappy sound everywhere...that's Bose.

Coincidentally (pun in tended) I downloaded the two Dahlquist patents that apply to the DQ-10, a speaker I have in my 2nd system. It's interesting to note that Dahlquist does not talk about matching driver positions by driver depth, as everyone assumed he did. Rather he states that rise time is the criteria for driver positioning. The DQ-10 has many issues, which I won't get into, but I wanted to see the theory behind the product. Heyser showed us that the true position of a driver in space is a continuum. MAS is correct that an active EQ with temporal correction is the only way to get close to solving the problem. I would add that a listener position locator, maybe a transponder on the listener, would allow a further refinement. The system could sense where the listener was and adjust the EQ and delay accordingly. Of course, what if you have more than one listener? We might be back to that famous New Yorker cartoon where stereo listeners are in a totem pole configuaration in the narrow sweet spot.

[DrWho]

I would add that a listener position locator, maybe a transponder on the listener, would allow a further refinement.

How bout a microphone at the listening position? []

[Guest srobak]

Just so I understand this in laymans terms...

You are trying to say that the radiation pattern... axis I guess... of not only a driver, but the perceived axis of an entire cabinet changes due to equalization? Would this be on a horizontal or vertical plane, or... ?

[DrWho]

I think mas is trying to describe a behavior in terms far far more complicated than it needs to be. Namely, if one uses EQ to fix the sound at a certain location, it is almost guaranteed that the EQ will introduce a new problem elsewhere in the room. Generally speaking, this is more of an issue in live sound where you have multiple listening positions. However, it's still an issue in the home, even for a single listening position scenario, because we don't listen with our head's in a vice.

The topic I originally brought up is that the radiation pattern changes at the crossover frequency based on the type of crossover and the time delay. Since crossovers are gradual reductions, you also have multiple frequencies around the crossover where the radiation pattern changes....and the actual radiation pattern is different for different frequencies. I believe Tom commented that the advantage of steeper crossovers is that there are less frequencies being overlapped, so the radiation patterns change over less frequencies.

[mas]

What Doc says regarding different crossover topologies (LR, Butterworth, Bessel,etc.), slopes, and signal delays (due to varying driver acoustic origin offsets) effect how the overlapping passbands interact. The superposition (combination) of these passbands results in comb filtering (in the frequency response) and polar lobing (in space - areas where the signls sum and are present in various intensities, and areas of out of phase interaction where frequencies are cancelled.

The topic of EQ was mentioned as an expansion on the principle. EQ has traditionally been the tool of choice in the attempt to correct this. But it fails to correct for errors resulting from the superposition of non-aligned (non-minimum phase) signals.

Ironically, what most often resulted in the apparent correction was that the components in a traditional EQ circuit emply variations of RLC circuits (resistors, inductors and capacitors). The inductors and capacitors modify the phase of the signal to a small degree. And in doing so, the signal alignment is altered slightly. By doing so, the orientation of the polar lobing and comb filtering frequencies present in the overlapping pass bands shifts slightly.

The significance of signal alignment and the combination of signals referred to as superposition, along with the resultant comb filtering and polar lobing is something that effects many 'levels' of audio, be it separate drivers (as in a speaker), multiple speakers, and/or room effects in the form of reflections. It is useful to become familiar with both the causes and the ramifications of this, as well as to understand how signal alignment in the time domain can be used to mitigate much of the destructive interference.