Ok so I've commented a few times on other threads about the notion that bass management in traditional HT products, a hold over from the THX legacy, is flawed. Based on some good science and some bad science, but ultimately focused on simplicity instead of optimal sound.
The current standard approach to bass management uses true crossovers, meaning there is a symmetric high-pass and low-pass. A 4th order LR low pass for the subwoofer and a 2nd order BW high pass for the mains (that when mixed with a sealed enclosure speaker with an 80hz -3db fs, would yield a 4th order LR high pass acoustically). This made sense if the mains were sitting on top of the subwoofer and/or crossed quite a bit higher. However, when we consider that the length of a period at 100hz is 10ms or 11.4 feet, what sense does it make to use symmetric and steep crossovers. The 80hz standard chosen for THX came from the fact that this was 2 standard deviations below the point at which people could not detect the location of a subwoofer. It was thus believed that this was sufficient to ensure that the majority of people could not detect the subwoofer's location.
Let's start by discussing the benefits to the standard approach, it has merit after all.
Advantages:
-It's easy for most people to understand and utilize, setup is simple, and good results are possible
-The high pass filter on the mains can improve their dynamic range, reduce the load on the main amp, and reduce distortion
-It's simple enough that automated setup algorithms can be made to work successfully with fewer rules to follow. My approach would require predictive modeling of the various scenarios to detect the optimal settings, something not possible a decade or two ago.
-I'm sure many of you will come up with many other benefits, as I said, it has its virtues, so those go here.
Before discussing the specific disadvantages, it's important to look at the limited research conducted here. The one I cite most for this is Earl Geddes, in the early 80's he completed a PhD Dissertation LF's in small rooms. At its core, the paper developed the first complete mathematical model of LF sources in small acoustic spaces along with the inclusion of the effect of boundaries. The models were confirmed through testing, but of course, the advantage of modeling is that we can also test a lot of scenarios that we wouldn't want to test in real life. His conclusions were interesting and cause us to want to draw conclusions in contrast to the traditional bass management approach, as well as the notion of 2-channel full range speakers. First, that there are specific locations that are best for subwoofers, and they are near barriers. Second, that damping on one wall per dimension is all that is needed (but doing both doesn't hurt, it is just not more effective than doubling it on just one side). Third, that bass smoothness was poor with traditional LF sources, such as 2-channel full range speakers or a single subwoofer. In fact, there was a direct correlation to the number of LF sources and bass smoothness. Spreading the LF sources out around the room activated more of the modes in the room causing there to be higher modal density and thus less variation in the peaks and dips. While it didn't necessarily make the response flatter or smoother, it did make it more consistent. We can eq consistent, but we can't eq variable.
Now, this work was theoretical, it did not explicitly test a subwoofer integration method. That didn't happen until 20+ years later with Welti, Toole, etc. at Harman. Their work copied much of the same principles, but in the context of an applicable approach (not that Geddes didn't develop his own, he just didn't publish the subwoofer method itself). Both Geddes and Welti's approaches are more similar than different, which I address in my Geddes Sub video. However, the key difference in Geddes approach, which matters here, is the bass management. Geddes notes that since the more LF sources the better the consistency of the bass, that the mains should be LF sources too. In fact, the modeling even suggested that this was a better scenario than simply adding more subwoofers. Integration and smoothness, as well as consistency, was improved when the mains were operated at full range (that doesn't mean response down to 20hz, just that they didn't have a high pass filter). Geddes also advocates for shallow low pass filters since steep filters don't really benefit us at such low frequencies, but do potentially cause harm in the sense that they increase group delay.
Disadvantages of the traditional approach:
-It neglects the mains as an additional LF source
-It increases group delay
-It can be less spatially robust in both consistency overall and at the integration point
-Speakers have different responses that make the use of asymmetric electrical filters with equal frequency dumb, that only made sense if all speakers were built the way THX indicated.
I think that last point may be one of the most important. Most speakers are ported and most have different port tunings from each other (that is, most speakers aren't tuned to a single common frequency, every speaker is different). Further, the overall response of the speakers varies greatly, so while two speakers could both have a port tuning of 35hz, one may be -6dB at that point and the other +3dB. As such, creating symmetric filters, as the original approach dictated, is impossible. Just like a speaker crossover often uses different slopes and different frequencies between, say, the tweeter and midbass to achieve a flat summed response, so should the subwoofer and mains.
Now remember something else, all speakers have an acoustic highpass. There is no such thing as a speaker with no highpass filter. The enclosure itself (be it ported, sealed, or TL) have a natural highpass. Since the shape of that highpass and its -3dB is different for every speaker, we need flexibility in the tuning of the high pass filters. Both its frequency and slope needs to be adjustable separately from that of the subwoofer low pass. Further, it should be standard on all devices that you simply have no highpass, but that this is unrelated to bass management (directing all bass from all channels to a dedicated bass/LFE channel.
My point with this is not to promote a brand or model, but the notion that bass management should be more flexible. That we hold back the ability to optimize systems with the current approach.
Unfortunately, the misunderstandings around low frequencies in small rooms and the way bass management actually operates leads to so much confusion that I fear there will never be a change.
I am not naive, I don't think most people will want to take the time to study and learn these concepts. Instead, I suspect that our saving grace will be advancements in bass management that will allow more nuanced decision making by the processor. If more LF sources is better, and this improvement is most notable above 50hz, then even fairly robust surround speakers could aid in the improvement of LF seat to seat consistency, smoothness, and bass integration.
Ok what about group delay, what is this business and how serious is it? Well, normally I would say not very important. I mean, the science is thin, but what we have suggests that if the group delay is kept below 1.5 periods, it shouldn't be audible. 100hz period is 10ms, that's a long time for a speaker, it would take a lot to add more than 15ms of groupd delay, right? Well, maybe not. The times are changing, and modern DSP does a lot to screw this up. First, the sub amplifiers now have DSP and DSP itself has overall delay. But a subwoofer is bandwidth limited. What happens when you add a bandwidth limited speaker to the overall system and it has a delay. It becomes group delay.
I will use WinISD to show a series of changes to the natural group delay of a single subwoofer progressing from no added filters to a worse case scenario where someone using the fanciest new DSP powered sub has gone to down with REW, attempting to fill in every little dip with boost EQ, added steep LP filters, etc.
No issues:
View attachment 27705
Just a low pass filter:
View attachment 27704
OMG What have you done!
View attachment 27703
Ok the last one isn't as bad as it could be, but for the sake of argument, after lots of filters, we are at or beyond 1 period across the entire bandwidth. While group delay was generally said to be a non-issue, they weren't thinking it could get this bad!
Now the room itself also gets in the way and so what I'm showing here is still better than reality. Modes add a ton of group delay, and I'll show some real-world examples of this. My point with all this, though, is that we should take steps to not introduce excessive amounts of extra group delay. I've now begun to measure peoples systems reporting more than 100ms of group delay, well over 30ms by 100hz! That isn't good.
Here is a measurement of a good system in room. This wavelet shows what we would want to see in a pretty ideal system. The black line reflects group delay, basically.
View attachment 27707
This shows a common setup with subwoofers in an ok arrangement. Not terrible, but not as good as above:
View attachment 27706
And...This shows the OMG what have you done scenario:
View attachment 27708
I think this is showing what happens when DSP runs wild. The solution is to go back to the DSP in the processor, handling the bass management, and provide the needed facilities to improve bass integration. This will hopefully reduce the reliance on DSP later in the chain (whose delay can't be compensated for if it is not common to all subwoofers) and where often far too much "fix" is applied.
When I setup a system and I "roll my own" bass management, by the time I've completed all of the setup work related to crossover, phase, speaker placement, etc., I only need 1-2 eq filters, and only cut filters. However, that is never possible with bass management the way it is.
