Improving Transition Range FR Swings in a 2.2 System

[klettermann]

So this is the same concept as placing subs around the room to fill nulls, but at higher frequency.

Exactly.

The reason it works with subs (and even this is somewhat controversial) is bass is hard to localise in-room. I've informally found that I can kinda hear direction down to 65hz or so. But in general it's not that big a trade-off to have bass sources at "wrong" locations.

In contrast, doing this at 200hz or anywhere near it will be IMO really distracting and will create a strange stereo image.

Yeah, though I want to confirm whether it's actually "will be" or "might be." So listening will answer that. If the benefit of FR smoothing is greater than some loss of directivity or sounds moving around by themselves, then that will decide.

I think this is one of those situations where acoustic treatment is probably a much better answer.

Probably, but I'm limited by various practicalities. I know for sure that moving the mains can help the trouble spots, but it's whack-a-mole cause it just creates other problems someplace else. The current positioning was very meticulously optimized, including imaging and soundstage, which are quite spectacular. The trouble zones are a small price to pay and the system sounds damn good anyway. I just want to see if I can do better!

 

[10PSPMemSticksForMiku2007]

You might be able to do this without ruining your stereo image if you can position your helper woofers in a direct line between your main speaker and listening position, and apply the correct digital delay.

 

[Flaesh]

the dual-sub setup. See below,

Just run two subs in an SSS configuration. Or better yet, DBA.Your room is long enough; have you 22' delay?. IMO and IME. (There can be other opinions, sometimes))
May be SSS is enough and there’s no need to place the subwoofers at a height of 3'7":

At higher frequencies, you can try optimizing the position(?) and regular absorbers(?). Although I like the "midbass helper" idea. But I don't know how that will sound in practice. Are you planning to use this only for the dips, with appropriate equalization for the "helper"?

ps: Or, conversely, you could try to hear that so called stereo bass...

 

[klettermann]

Just run two subs in an SSS configuration. Or better yet, DBA.Your room is long enough; have you 22' delay?. IMO and IME. (There can be other opinions, sometimes))
May be SSS is enough and there’s no need to place the subwoofers at a height of 3'7":
View attachment 557905
At higher frequencies, you can try optimizing the position(?) and regular absorbers(?). Although I like the "midbass helper" idea. But I don't know how that will sound in practice. Are you planning to use this only for the dips, with appropriate equalization for the "helper"?

ps: Or, conversely, you could try to hear that so called stereo bass...

@Flaesh Thanks for taking the time to model this. I had to look up SSS—my terminology is still catching up.

Anyway, I fully appreciate the potential benefits of different sub placements, including whatu’ve illustrated. Unfortunately, the front and rear wall midpoint locations simply aren’t available in this room. I should probably have emphasized that practical constraint more clearly. Part of the motivation for the helper experiment is to see what’s achievable with the main speakers and subs in their existing locations.

And yes, you’ve understood the helper idea correctly: band-limit and EQ its output so that it contributes primarily in the deficient regions, with little contribution where the existing response is already satisfactory. Then adjust level, delay and polarity and measure the combined result. Whether that actually produces a useful improvement, and how it sounds, is what I hope to find out. I haven’t yet properly tested it, so this is an experiment rather than a demonstrated solution. Cheers,

 

[Flaesh]

SSS

https://www.academia.edu/5362839/Perception_Modal_Control

https://www.audiosciencereview.com/...-fidelity-home-reproduction.1259/post-1328182 ))
It’s simply a degraded DBA\CABS, but in some cases, that degradation might be negligible compared to the "proper layout."

the front and rear wall midpoint locations simply aren’t available in this room

Are they inaccessible at any height, right up to the ceiling?
In O800 ARAM manual (attached) a really strange(?) idea is being proposed for the two subs. You may play RoomSim or try real experiment.
Travelling plain wave undoubtedly differs from the combination of resonances and other reflections. But how well will it work in a real room? And how do we (they, you, someone) hear it?))

you’ve understood the helper idea correctly: band-limit and EQ its output so that it contributes primarily in the deficient regions

Here is another approach:

But it might not be particularly useful in your case. The nulls are becoming deeps. And the time-domain performance improves.

 

[klettermann]

An overdue update: I’ve now worked through a more systematic version of the helper experiment and finally spent some time listening. To my considerable surprise - and delight!! - the improvement is not subtle.

The concern raised earlier about stereo imaging was also my biggest reservation. So I’ll start there: I cannot hear a loss of separation. If anything, things sound better anchored. More on that below.

This follows the fresh start described in post #17. For this stage I isolated the Maggies and helper, with the subs off. The helper is still the single SMGa in the rather unconventional position described earlier - horizontal, centered against the front wall behind a large TV and angled upward at about 45deg. Processing remains in the miniDSP hardware. First, a reminder of the problem:

Chart 1: Initial Responses

Orange is left, green is right, and blue is L+R. The individual trouble spots around 145 Hz on the right and 233 Hz on the left carry through quite obviously into the combined response. There is another persistent depression around 380 Hz. Why not just EQ the mains better? I tried. Within the placement and treatment constraints already discussed, this was the best compromise I could achieve without making something else worse. That was the reason for trying another acoustic source in the first place. Anyway, the system did not sound bad or somehow deficient, damn good in fact despite the flaws. The process, in outline:

1. Measure the mains individually and together, and the helper separately, using a shared timing reference.
2. Band-limit the helper, initially with LR48 filters at 100 and 600 Hz.
3. Generate a band-limited inverse-magnitude curve from the L+R mains response as a starting target for the helper.
4. EQ the helper’s measured response toward that target.
5. Load the filters and remeasure the helper to verify the result.
6. Adjust relative delay and level against the actual combined response.
7. Refine the upper cutoff, then revisit delay and level.

The distinction in step 4 matters: the inverse curve describes the desired acoustic output of the helper, rather than simply being a filter applied to it. The helper has its own rather uneven room response that must first be taken into account. And, as discussed earlier, complementary magnitudes alone don’t guarantee a useful sum; phase and level still have to be worked out.

Chart 2 — Shaping the helper

Red is the crossover-filtered but otherwise unEQ’d helper, blue is the target derived from the mains, and green is the measured EQ’d helper. This illustrates the shaping stage before the final integration adjustments. REW’s automatic EQ got me partway there. Manual adjustments gave a closer fit, followed by quite a few measurements of different delay, level and crossover settings.

The current working combination is a 100–550 Hz helper band, LR48 at both ends, 4.0 ms delay on the mains, and a helper gain setting of −7 dB. The gain figure is just the setting in my particular chain, not an acoustic level relative to the mains.

Chart 3 — Mains alone versus mains plus helper

Blue is L+R alone; red includes the helper with the selected timing and level. The traces are level-matched for comparison. All three charts use 1/6-octave smoothing. It plainly isn’t perfect. The depressions are reduced, but some peaks increase, and the remaining notch around 240 Hz has resisted improvement without unwelcome tradeoffs elsewhere. Nevertheless, the broad deficit through the upper bass/lower midrange is substantially filled.

Now the interesting part: listening. Earlier I said I wanted to find out whether the imaging problem “will” happen or “might” happen. In this particular setup, I haven’t heard the any of feared degradation! In fact, the presentation actually sounds more spatially balanced. The center is better-defined and moves farther forward. Going back makes the center comparatively recessed and the overall presentation somewhat thinner.

There is a little more warmth, which I like, but the impression goes beyond warmth. Voices and instruments simply seem better anchored. That was not the benefit I set out to obtain, and I certainly wasn’t expecting it from adding a mono helper.

These are informal listening impressions, not a blind test or a claim that the measurements explain every perceived change. But the audible benefit is substantial enough that the experiment has been worth the effort (which was actually considerable). This also changes my attitude toward the proposed stereo LS50 helpers. I may still try them, but they now have something worthwhile to beat. I don’t want to give up the center presentation I’m hearing just to obtain a tidier response plot.

So, for now, I’m going to live with this version. Integration with the existing mains/sub system, nearby-position checks and further examination of time-domain behavior remain to be done. But this is already a very useful result at the listening position. And finally, the effect isn't some head-in-a-vise solution. It all held up on each side of the MLP - another nice surprise.

Thanks to everyone who contributed ideas and raised the imaging question. That actually turned out to be the most surprising part of the project! Cheers,