So, time alignment is kind of important.

[CB89]

Until now, I kept things simple regarding delay and phase: subwoofer at the front, between the speakers, phase at 0°, and done.

Curiosity got the better of me, so I connected a hardware DSP to a secondary system (desktop, near-field setup) specifically to enable independent delay adjustment for the speakers and subwoofer. I also did some reading on the subject and ran a test using REW. I performed separate measurements for the left and right channels and the subwoofer - using the "Acoustic time reference" feature - then used the "Time Alignment Tool" to determine the optimal delay settings and entered them into the DSP.

The difference was surprisingly significant. Even though the speaker delay was only 3.9 ms - and despite the need to invert the subwoofer - the soundstage changed dramatically. The perceived amount of low-end is significantly reduced, yet what remains is crisp and precise. It now sounds more like a 2.0 system with extended bass response than a sub/sat system.

Paradoxically, this runs counter to the measurement results. By correcting the delay, there are fewer dips in the crossover region - meaning, in effect, more low-frequency output.



I explain this to myself by assuming that bass impulses were previously "stretched out," creating the impression that there was more low-end present before.

So that from "less" measurable low frequencies - here symbolized by two circles of different sizes:


..which, subjectively, was more present before, just stretched out and more washed-out.




For my main system, I have now acquired a hardware DSP as well and was able to reproduce the result there, too. Interestingly, with similar values (4ms delay, inverted subwoofer). Admittedly, it is the same room, but with midfield positioning, different loudspeakers, and a different subwoofer.

For the sake of completeness: the equalized frequency response (without target curve).



I was under the impression that I already had a well-balanced 2.1 system, but this marked a significant leap forward.

 

[kyuu]

I explain this to myself by assuming that bass impulses were previously "stretched out," creating the impression that there was more low-end present before.

Nothing you've presented supports this assertion. The frequency response changed, which does explain your different perception of the sound (along with confirmation bias).

 

[thcdru2k]

Proper delay/polarity improves acoustic summation and reduces cancellation through the crossover region. This improves bass quality as your speakers and subs behave more like a single source making the crossover region harder to localize. Your graphs show clearly the improvement from 65hz to 95hz

 

[CB89]

Proper delay/polarity improves acoustic summation and reduces cancellation through the crossover region. This improves bass quality as your speakers and subs behave more like a single source making the crossover region harder to localize. Your graphs show clearly the improvement from 65hz to 95hz

It does not manifest quite so drastically (after equalization). For comparison, here is an older measurement without time alignment.

 

[CB89]

Nothing you've presented supports this assertion. The frequency response changed, which does explain your different perception of the sound (along with confirmation bias).

This is what Gemini says about my impression. How would you classify this response?

What you are describing is the difference between temporal coherence and mere frequency response optimization. The fact that the bass now sounds "leaner" is not a sign of loss, but rather of precision.

Here is the explanation as to why your hearing perceives a much greater change than the frequency response graph would suggest:

1. The End of "Ringing"

When a subwoofer and loudspeakers are not perfectly aligned in time, phase shifts occur. These shifts cause the sound waves in the crossover region (your 80 Hz point) to arrive at your ear at different times.

Before (Muddy): The subwoofer lagged behind the loudspeakers. The signal became "smeared" in time. Your brain interpreted this temporal delay as a prolonged, imprecise signal—the typical "boomy" or "slow" bass.

Now (Punchy): All frequency components of a kick drum reach you at the exact same moment. The attack (the "kick") and the body of the bass merge into a single, unified impulse.

2. The "Comb Filter Effect" in the Time Domain

A smooth frequency response graph in REW merely indicates that the total energy is correct. However, it reveals very little about how that energy is distributed over time.

Psychoacoustics: Our hearing is extremely sensitive to the rise time of signals. If the phase alignment in the crossover region is incorrect, the group delay becomes erratic. This causes certain frequencies to linger in the room slightly longer than others. Correcting the delay has straightened out this group delay.

3. "Lean" Usually Means "Correct"

Many users become accustomed to the "fat" sound of a poorly integrated subwoofer because phase errors often create an artificial boost or a sense of "dragging"—qualities we mistakenly interpret as simply being "lots of bass." The reality: True, precise bass is dry. It is present only when the signal calls for it, and vanishes immediately thereafter. At first, this may feel like "less"—but technically speaking, it constitutes a faithful reproduction.

 

[thcdru2k]

It does not manifest quite so drastically (after equalization). For comparison, here is an older measurement without time alignment.
View attachment 532301

I personally do not like heavy handed EQ but to each their own. I just like to cut some of the peaks no boost. I rather decrease sub level if crossover region is too much of a hump then use EQ.

 

[Bjorn]

Now you understand how ridiculous passive speakers are with no time alignment between drivers. The improvement is also higher closer to the presence area.

 

[kemmler3D]

This is what Gemini says about my impression. How would you classify this response?

Typical of AI responses that seem to have been trained on Audiogon threads, it names a bunch of real concepts, with more-or-less correct descriptions, but misapplies them to the situation.

Look at the waterfall or other views in REW that show how long the sound is decaying to get a sense of how correct this Gemini response might be.

A difference of +/-4ms might have various audible effects, but the decay time is mostly determined by the room, and will be about 20-50x longer than the change in timing you introduced.

 

[Keith_W]

The difference was surprisingly significant. Even though the speaker delay was only 3.9 ms - and despite the need to invert the subwoofer - the soundstage changed dramatically. The perceived amount of low-end is significantly reduced, yet what remains is crisp and precise. It now sounds more like a 2.0 system with extended bass response than a sub/sat system.

A 3.9ms delay is well under published audibility thresholds for group delay at low frequencies. Studies done with test signals and headphones indicate that the audibility threshold is at least one period of the frequency in question. So if we are talking 80Hz, one period is 12.5ms. This means you should not be able to hear it, provided the frequency response does not change.

That the study was done with headphones is important here. Changing the group delay with an all-pass filter in a headphone will not alter the frequency response. But as you can see, adjusting the delay of the sub with DSP does alter the freq response.

This does not mean you shouldn't do it, but we should be realistic about attributing what we hear to what we changed. Confirmation bias can be very strong, I convinced myself for a long time that I could hear a 1ms delay in my subwoofer. That was before I read that study, and suddenly I can't hear a 1ms misalignment any more.

 

[thcdru2k]

One advantage of passive speakers with external DSP is modularity. You can optimize subwoofer integration, crossover behavior, delays, routing, and room interaction without being locked into a fixed internal DSP/amplifier architecture. It also makes upgrades easier since speakers, amps, DSP, and subs can all evolve independently.

 

[Chris A]

That was before I read that study, and suddenly I can't hear a 1ms misalignment any more.

Nice punch line...!

 

[CB89]

I personally do not like heavy handed EQ but to each their own. I just like to cut some of the peaks no boost. I rather decrease sub level if crossover region is too much of a hump then use EQ.

I’m not sure what you mean; we’re probably talking past each other. I don’t boost any frequencies; I simply use 2–3 PEQs per channel in the 20–300 Hz range. That isn’t the point, anyway; the point is that—thanks to time alignment—the low end sounds (subjectively) leaner yet punchier, even though there are now fewer dips in the crossover region, resulting in a higher overall low-frequency level.

 

[CB89]

Look at the waterfall or other views in REW that show how long the sound is decaying to get a sense of how correct this Gemini response might be.

A difference of +/-4ms might have various audible effects, but the decay time is mostly determined by the room, and will be about 20-50x longer than the change in timing you introduced.

Yes, I'll check later to see how the change is reflected in the waterfall. It probably won't be particularly significant. Nevertheless, I'm curious where that impression comes from.

 

[Redacted]

Sound travels about a foot in a millisecond.
I always put my subs on my knees to ensure perfection.

 

[CB89]

A 3.9ms delay is well under published audibility thresholds for group delay at low frequencies. Studies done with test signals and headphones indicate that the audibility threshold is at least one period of the frequency in question. So if we are talking 80Hz, one period is 12.5ms. This means you should not be able to hear it, provided the frequency response does not change.

That the study was done with headphones is important here. Changing the group delay with an all-pass filter in a headphone will not alter the frequency response. But as you can see, adjusting the delay of the sub with DSP does alter the freq response.

This does not mean you shouldn't do it, but we should be realistic about attributing what we hear to what we changed. Confirmation bias can be very strong, I convinced myself for a long time that I could hear a 1ms delay in my subwoofer. That was before I read that study, and suddenly I can't hear a 1ms misalignment any more.

Yes, I would certainly like to get to the bottom of that. There are a few factors that could be at play here:

- A different crossover slope resulting from the switch from the subwoofer's internal crossover to the DSP.
- Constructive summation of frequencies within the crossover region (80 Hz)—and the surrounding frequencies—due to the delay settings, compared to before.
- Inverted phase on the subwoofer

And just as a side note: a different DAC (switching from an audio interface to a miniDSP Flex). However, I don't suspect this caused any change; I just wanted to mention it.

One area where even minor changes in delay can have a significant impact, though, is the phantom center. A reduction of just 2 ms on one speaker can cause the phantom center to shift noticeably. I assume, however, that this isn't what you were referring to, and that you were instead referring to group delay. I mention this point simply to highlight the fact that small changes in delay can indeed make themselves felt quite distinctly in other areas.

Regarding "confirmation bias," I plan to conduct a few A/B comparisons using different presets in order to verify my impressions. To do this, however, I would need to revert to the state where I perceived the bass as being slightly boomy.