How important in phase control on a subwoofer?

[KSTR]

A quick simulation.
Assumed is a THX-compliant Sub+Main arrangement:
- Sub is acoustic lowpass 4th order Linkwitz Riley ("LR4) at 80Hz (or 85Hz, forgot exact number).
- Sub provides a highpass output for the mains, Butterworth 2nd order at 80Hz. This must also compensate the acoustic highpass function of the sub, say 20Hz 4th order (sealed sub with subsonic filter, or a ported box sans filter).
- Mains are acoustic highpass 2nd order Buttherworth at 80Hz, and this combines with the electrical highpass for the needed total filter, LR4 again
This sums perfectly:

I actually modelled a 3-way with 300Hz 2nd order XO woofer to mid, this adds some low frequency group delay which in this case happens to be exactly compensated by moving the sub back by a millisecond (30cm).


Now let's increase the time-of-flight difference to 3.5ms having the woofer even further back a bit above one meter. Even with a polarity switch we cannot compensate the dip:

Adding in a delay (phase shifter) on the sub, we can arrive at this:

Not excactly perfect but better than without.

But it's not going to sound good, the low bass will be severly lagging.
Looking at group delay we can see why:

At 80 Hz the sub is now way late compared to the main. The step response looks equally bad:

Not that the original step response were exactly what one would call "quick", yet it is still significantly less ringy:

Now, mind you, we started with a perfectly aligned system... in any actual use case, notably with generic subs and mains, the chances of what you get is pretty much random. There was a good reason for setting up the THX specs as it does help to reduce the unknowns so the customers are not competely left in the dark.
BTW, overall magnitude and phase compensation with DSP to "fix" the total response at the listening position does not work well when the system is heavily "detuned" already. It's gonna sound better, but not really as good as it could be. You have to have the basics right, and thats the dilemma that I mentioned. Same goes for adjustable delay for the mains, optional filters, etc in AVRs, this can fix only so much (though often better than with analog means alone).

 

[MZKM]

Right, most subs actually use phase and not a linear delay, PSA and Rythmik being exceptions.

SVS also mentions delay (states not to use it if you have an AVR).

Do you know any others besides JL that do a filter?

 

[Ron Texas]

Yes measure for sure it would be darn impossible to dial in by ear only especially with music . I used the built in DRC in my pre pro and dialed in the distances etc , but still i was left with some options to decide between .
The aforementioned increments in distances for example my actual distance is between two notches so which to choose ? REW and umik to the rescue

I don't think it has to be perfect, but I don't know what the acceptable range is.

 

[Mnyb]

I don't think it has to be perfect, but I don't know what the acceptable range is.

Not me either i don’t know thats why a measurment can help you see the effect play out . I’m no expert there is a lot of trial and error for me.

 

[Chromatischism]

SVS also mentions delay (states not to use it if you have an AVR).

Yeah, good point - I think most DSP systems would use a delay. Rythmik doesn't use DSP but has another way of implementing a true delay. Brian explained it at AVS once upon a time. You can see the delay markings on the sub dial, similar to PSA.

HSU has a 0/180 switch which indicates a traditional phase control, and I think that should be a safe assumption for most subs with that marking. If anyone is making subs with a 0/180 deg marking but is actually using a DSP delay, shame on them for confusing the customer. I don't think it's that many though as that would be a feature worth showcasing.

 

[dasdoing]

btw: if there is only a 0 and 180 degrees switch it is no phase shift at all, but a polarity reversal.

The mains need to be delayed in relation to the subwoofers

that is the problem. for music the phase controls on subwoofers are useless. it is ok for movies to have aligned phase 360 degrees behind I guess

 

[Chromatischism]

btw: if there is only a 0 and 180 degrees switch it is no phase shift at all, but a polarity reversal.

potayto, potahto

 

[warthor]

I wonder if Erin would be willing to chime in on this issue since I have heard him bring it up on numerous occasions. @hardisj ?

 

[hardisj]

I'm late to the party but to address a few things:

In car audio and pro audio DSPs I have first hand experience with, they incorporate a "variable phase" control which is a 2nd order all pass filter.
This is targeted to a particular frequency (usually the LPF). And this IS NOT standard time delay. They are completely separate and serve as two entirely different functions.

In the simplest of terms, time delay serves to delay the entire signal; all frequencies. An all-pass filter is targeted to rotate the phase angle at a particular frequency.

Given this experience with other DSPs, I was under the guise that the home audio subwoofers with "variable phase" were just that: variable phase. However, and unfortunately, that has shown to not be the case. They are instead just glorified time delay knobs. Which, IMHO, sucks.

Why does it suck? Variable phase is a crucial element to getting subwoofers and mains blended properly. From personal experience (again), I can tell you this is a huge benefit. You can easily have two speakers be in time but be out of phase. It's a simple concept to grasp. Just put two speakers side by side, and flip the polarity. There you go: same IR time of arrival but 180˚ out of phase. Therefore, you can imagine that it is completely possible (and likely) you will have speakers arrive in time but still out of phase at the crossover region between a subwoofer and the mains. You can be as little as 10˚ or as much as 280˚ out of phase between speakers at their crossover region. As you can also imagine, correcting the phase alignment will increase coherency (and is completely measurable). This is more audibly noticeable in LF; less in HF.


I hope this is "enough" because, frankly, I don't know how else to convey in words just what this means to a sound system without providing an in-situ demonstration. Something I have done numerous times over the years in car audio. When you hear it you understand just why it is important.

 

[KSTR]

In car audio and pro audio DSPs I have first hand experience with, they incorporate a "variable phase" control which is a 2nd order all pass filter.
This is targeted to a particular frequency (usually the LPF). And this IS NOT standard time delay.
In the simplest of terms, time delay serves to delay the entire signal; all frequencies. An all-pass filter is targeted to rotate the phase angle at a particular frequency.

Well, actually it is a true time delay, up to the corner frequency of the group delay. Of course you can have different shapes of the group delay around cutoff, depending on Q factor... which is normally choosen to get a maximally flat delay up to the corner (Butterworth alignment of the group delay) which results in Q=~0.6, the same value that would give a Bessel lowpass response of magnitude. Cascaded 1st order is sometimes used, with resulting Q=0.5 and a somehwat shallower transition of the group delay into its roll-off.
I haven't seen any subwoofer phase shifter that doesn't use 2nd order delay allpass with adjustable group delay corner frequency, and Q in the 0.5 to 0.6 range, which results in variable time delay (the higher the cutoff the shorter the delay).
EDIT: It is true, though, that once the delay can be made so large that the corner frequency gets close to the XO frequency then the thing stops being a pure time delay and the phase added to the sub channel isn't corresponding exactly to a time delay but it doesn't matter as long as the needed phase at/around the XO can be dialed in.
But overall as noted, phase adjust is a not so well working bandaid .

 

[JSmith]

potayto, potahto

… tomayto, tomahto, let's call the whole thing off.

Sub crawl... not to be confused with pub crawl.



JSmith

 

[dasdoing]

potayto, potahto

I researched and it means "same thing"?

flipping polarity wont fix time misalignment. you can fix some cancelation, sure....but not the group delay

 

[GalZohar]

I still don't fully understand what those phase adjustments are doing (and which hardware contains which kind of adjustments).

The practical question I have is basically if there is any useful purpose to the phase knob on the SVS pb-2000 over just adjusting delay in the AVR?

And in Anthem receivers, is there any useful purpose for the subwoofer phase adjustment and phase adjustment center frequency over just adjusting the delay?

If I had some means to apply any filters (such as all-pass) to my main speakers, could I improve my subwoofer integration without messing something else up (such as imaging) on the way?

I read a lot of technical explanations, but my ability to understand them is not good enough in order to get the above questions answered.

 

[GalZohar]

A quick simulation.
Assumed is a THX-compliant Sub+Main arrangement:
- Sub is acoustic lowpass 4th order Linkwitz Riley ("LR4) at 80Hz (or 85Hz, forgot exact number).
- Sub provides a highpass output for the mains, Butterworth 2nd order at 80Hz. This must also compensate the acoustic highpass function of the sub, say 20Hz 4th order (sealed sub with subsonic filter, or a ported box sans filter).
- Mains are acoustic highpass 2nd order Buttherworth at 80Hz, and this combines with the electrical highpass for the needed total filter, LR4 again
This sums perfectly:
View attachment 115881
I actually modelled a 3-way with 300Hz 2nd order XO woofer to mid, this adds some low frequency group delay which in this case happens to be exactly compensated by moving the sub back by a millisecond (30cm).


Now let's increase the time-of-flight difference to 3.5ms having the woofer even further back a bit above one meter. Even with a polarity switch we cannot compensate the dip:
View attachment 115883


Adding in a delay (phase shifter) on the sub, we can arrive at this:
View attachment 115884
Not excactly perfect but better than without.

But it's not going to sound good, the low bass will be severly lagging.
Looking at group delay we can see why:
View attachment 115885
At 80 Hz the sub is now way late compared to the main. The step response looks equally bad:
View attachment 115887
Not that the original step response were exactly what one would call "quick", yet it is still significantly less ringy:
View attachment 115888

Now, mind you, we started with a perfectly aligned system... in any actual use case, notably with generic subs and mains, the chances of what you get is pretty much random. There was a good reason for setting up the THX specs as it does help to reduce the unknowns so the customers are not competely left in the dark.
BTW, overall magnitude and phase compensation with DSP to "fix" the total response at the listening position does not work well when the system is heavily "detuned" already. It's gonna sound better, but not really as good as it could be. You have to have the basics right, and thats the dilemma that I mentioned. Same goes for adjustable delay for the mains, optional filters, etc in AVRs, this can fix only so much (though often better than with analog means alone).

Supposedly if we delayed the mains instead (increasing subwoofer distance in the AVR), in this ideal theoretical scenario, wouldn't that solve everything?

 

[KSTR]

Supposedly if we delayed the mains instead (increasing subwoofer distance in the AVR), in this ideal theoretical scenario, wouldn't that solve everything?

Exactly. To get the proper integration with a THX sub, it must be placed close to the mains wrt distance. The Linkwitz-Riley alignment already is the most tolerant to distance offsets but once your approaching 1/4th wavelength things fall apart and the theoretically correct solution is to delay the speaker that arrives first no matter if it's the mains or the sub(s).

OTOH, placing subs and mains further apart may have benefits sometimes because that can create some directionality around the XO frequency which can help to reduce room mode excitation. For the same reason, nominally perfect integration of mains and sub may or may not sound better than a somewhat misaligned setup when you happen to have room modes at the wrong spots which now see different excitation conditions.

It's all about compromises notably when no DRC is applied. No silver bullet here.

 

[GalZohar]

I had to delay my speakers even more than the delay "measured" by Audyssey (increasing subwoofer distance). Optimal distance for left speaker was similar to the Audyssey distance, but for the right speaker I had to add 1.5m, which messed up the left speaker. To get the best of both worlds (almost as good as the best each can get separately) I had to add a total of 2m to the subwoofer distance over Audyssey value. Of course that gave flattest response, I have no idea if I should be looking at anything else as well.

I tried seeing if the phase knob on the subwoofer does anything but I couldn't figure it out from the measurements. It's still unclear to me exactly what it does on the SVS PB-2000, and if adjusting it can provide any benefit over just adjusting distance.

 

[ernestcarl]

I had to delay my speakers even more than the delay "measured" by Audyssey (increasing subwoofer distance). Optimal distance for left speaker was similar to the Audyssey distance, but for the right speaker I had to add 1.5m, which messed up the left speaker. To get the best of both worlds (almost as good as the best each can get separately) I had to add a total of 2m to the subwoofer distance over Audyssey value. Of course that gave flattest response, I have no idea if I should be looking at anything else as well.

I tried seeing if the phase knob on the subwoofer does anything but I couldn't figure it out from the measurements. It's still unclear to me exactly what it does on the SVS PB-2000, and if adjusting it can provide any benefit over just adjusting distance.

Dunno about the SVS, but the variable delay/phase knob in my Rythmik sub acts like an all-pass filter where it shifts the phase response (without affecting the frequency magnitude) like what a HPF normally would do -- I've observed that there's still some visible, but very minimal, actual shift in overall time delay, say ~2-3ms when rotated at max "16ms"/180 degrees.

I adjusted the step response peaks to be similar in the graphs below as measured at my couch MLP:

sub is already equalized for MLP

You can see that there's an increase in the GD. I would rather not use it personally, and would apply digital filters instead.

 

[dasdoing]

all-pass filter where it shifts the phase response

naaah. it's just a delay, like all variable "phase" knobs. calling those knobs phase-anything is just a stupid idea someone once had and more stupid people copied.

 

[egellings]

A phase angle between two signals from the same signal source indicates a delay between the two.

 

[dasdoing]

a pure impulse delayed by 16ms

results in a straight fall of the phase

notice though that I toggled to linear frequency axis.
on the log axis we are using most of the time it looks like this