Bass system integration

[Kvalsvoll]

How to integrate the bass-system/subwoofers with the main speakers, so that the end result actually sounds better. 

This involves setting crossover frequency and delay and level, so that the bass system and main speakers integrate seamlessly. Get those 3 steps right, and the result will likely have significantly improved sound quality compared to switching off the subwoofers.

Additional steps includes dsp processing - like eq - on bass system, dsp/eq on main speakers, placement of main speakers and bass system, room acoustics, placement of listening position. A complex process with so many variables. Kind of nirvana for the hobby enthusiast looking for problems to spend time on, a real pain for someone who just bought this awesome bass system and now want to hear some music and watch a movie.

Integrating the bass system is like engineering a speaker system in your room, which requires knowledge in electroacoustics and equipment to do acoustic measurements. So , what do you do if you have neither?

You use a set of rules to make the necessary settings. Perhaps helped by automated set-up functionality, such as audyssey. The result is what one can classify as "variable success" - in some cases, it ends up with something that sounds reasonably good, in other cases it is so flawed that the sound is better when subwoofers are disabled.

Even for experts, a set of rules makes it possible to do a system in a reasonable time frame, and also give more predictable results.

So, which set of rules applies. How do we set this up, simple and easy, with predictable results. Share your way of doing the integration, ask questions. Let us kill the myth that subwoofers must be switched off for music.

This thread hopefully will provide useful information on how to integrate a bass system, and when we enter the more advanced parts we may be able to further improve the methods we use.

[Kvalsvoll]

Integration basics:

We start with the simple basics that everyone on this forum already knows. And then we will see that what should be quite simple and straight forward, just is not so, at least not always.

Assuming the subwoofers are in place, and any eq and other settings on the bass system is completed. If it isn't perfect - well, it never is - then that's what we have.

The end goal is to end up with a system with a smooth frequency response matching our chosen target and very good transient behavior. Transients means timing - we need to get the time domain right, it is not enough to just eq or whatever to get a perfect frequency response.

To achieve this, we can adjust the level on the bass system to match the mains, we set delay on the mains so that the bass system and main speakers sum up correctly in the time domain, and we set a low pass filter on the bass and high pass filter on the mains - this is the crossover.

The typical processor allows tuning of these parameters.

 

Level:

This is where the typical buyer fails first - level is set too high, often 20dB higher on the bass system, because the subwoofer was expensive and must be heard.

The frequency response measurement effectively reveals this problem, simply adjust the level so that the frequency response matches your target as close as possible. Now if both mains and bass system are smooth/flat, this is quite easy, but can be more tricky to find the best compromise if the responses are very nonlinear. 

Level can usually also be adjusted on the bass system amplifier, so you don't need a typical processor for this.

 

Delay on mains:

This is the subwoofer distance setting. This should not be set to the distance to the subwoofers, it must be set to match the total delay in the bass system, which includes dsp, additional filtering, acoustic delay in subwoofers, acoustic delay due to room interaction. It will usually be a lot more than the physical distance.

One method to get it right is to look at the frequency response around the crossover - the correct setting is when the bass system and mains sum up to a reasonably flat level with no dips. The problem is that there are an endless number of distances that will look right in the crossover region, because if the delay is exactly one period wrong, it will still sum up. Fortunately the frequencies involved are quite low, so the delay for one period at crossover is long, and we can see this on the time step response or in the spectrogram. Tune it approximately first, using spectrogram and time step, then fine tune using frequency response.

Have an expensive audiophile preamp with no delay for mains? Well, why do you have that, sell it and get gear that is possible to set up properly to get good sound.

 

Crossover:

Choice of crossover frequency depends on main speaker response and capacity, bass system response and capabilites in the upper bass range, location of subwoofer units.

Generally, a higher cf gives better results, 120hz is often a good starting point to try. If your bass system has asymmetrically located units, a lower cf is needed to avoid localization, try 80hz, in bad situations as low as 60hz can be necessary.

It is also possible to run the mains with no high pass, running full range. If the main speakers have extension and capacity at low frequencies, they will now act as part of the bass system, hopefully resulting in a smoother total response.

 

All this is quite easy to do right if the bass system and mains have flat and perfect responses. In real world scenarios, this is not the case, and decisions has to be made to choose the best compromises. It is quite clear that the set up an calibration of the bass system before integration is also a part of the integration process, and then this is no longer easy to do just by following a few rules.

[Ukko Kari]

I concur that the time portion of the system response is what we must strive for correctness. Time of flight,  processing delay needs to be accounted for.

What I have done: in a multi-sub system, measure time of flight ( with no DSP engaged ) from each device in it's intended location to the main listening position, and compare that to the main L/C/R channels.

I helped set up a system with subs that were radically different in distance, with a nearfield sub behind the MLP area. In that particular case, the closest subwoofer was set for level, and run at a bandwidth of <40 hz. Delay was set so it's time of flight was identical to the average of the main L/R speakers. The other 3 subwoofers in this large room were set so their time of flight was identical. The pair closest to the L/R were run <120hz ( main speakers were not on the level of some enthusiasts here ) The third subwoofer was run <60 hz.

Subwoofers were not gain matched, the one behind the MLP was reduced considerably in level, it's 4kw amp was just idling barely with respect for the required levels.

Processing added 1-1.5 msec of delay, so the main and subs required a bit of fine tuning. Once all 4 were playing 'nice' with respect to each other and the main L/R, Anthem room correction was run.

Every situation is unique, and requires a customized solution.

You may get better smoothing in the upper bass running all subs as full bandwidth, but may result in a softening of the attack across some of the listening positions, if there are many, due to the various complex phase relationships in different physical locations. Longer wavelengths ensure that you will have better success at integration.

It's not hard to get sources within 1/4 wavelength at 20 hz, or even at 40.

 

[Kvalsvoll]

Yes, timing is the key.

Good idea to use lower cut-off for rear subs in a multisub system, it is not easy to achieve a very good result with sub units both up front and back.

[SME]

Hey.  I'm glad someone started this thread, even though I haven't really had time to post.

I generally agree with what's posted above as far as what to do as a starting point.  However, I personally have found the sub/main crossover (XO) issue to be ugly at best.  The trouble is that the LR4 XO, which is typically used is largely theoretical for subs/mains but not practically realizable.

In a speaker, an LR4 XO is typically implemented between drivers that are closely co-located and typically vertically-oriented.  For drivers that are direct radiating and installed flush with the baffle, the sound will typically be time-aligned and the XO will integrate ideally for listeners located anywhere perpendicular to the vertical axis of the baffle.  Also much of the off-axis sound that is heard, that which arrives from the sides, should also integrate ideally.

The ideal LR4 has each driver -6 dB from its pass-band level at the XO point.  Thus, the levels sum to 0 dB under ideal circumstances.  Under less-than-ideal circumstances, such as when the drivers are not time-aligned (a common problem with horn speaker design) or the listener off-axis vertically, circumstances are less-than-ideal, and the levels sum to less-than 0 dB.  For some XO types other than LR4, it's possible for the levels to sum to greater-than 0 dB, but this is often undesired as this peak in the off-axis spectrum will tend to be audible.

===

Unfortunately when it comes to subs, the different drivers involved in the XO are frequently neither co-located nor vertically oriented.  Furthermore, the wavelengths in the XO area are long enough for nearby boundaries to substantially interfere with the direct sound from each driver.  The consequence is that the LR4 XO for mains and subs is rarely ideal.  Using the "sub distance" setting, it is possible to make the LR4 XO ideal, but only for a single input channel (i.e. front center) for a narrow band of frequencies (rather than, necessarily the full XO range), and only at one seat location.

With more effort, other compromises are possible taking into account more than one channel and/or more than one seat, but it's almost never possible to achieve ideal results across the XO frequency range for multiple channels and multiple seats.  In fact, to do so practically requires co-location of the speaker and its own sub(s), which completely defeats the main advantage of bass management in the first place.  (I.e., subs in locations other than the speakers often perform better in-room well below the XO frequency.)

As above, where the XO circumstances are less-than-ideal, the response sums to less-than 0 dB, so without additional EQ, sub systems almost always exhibit a substantial power response dip or hole in the XO region, which happens to be a critical region for musical bass and for a lot of punch, impact, and chest slam feeling.  Considering that lack of punchyness or chest slam is a very common complaint for home systems, I suspect that sub/main XO problems are often detrimental to this aspect of the sound.  The only solutions are to fill it in with EQ, possibly causing an annoying resonance at certain seats and/or room locations,, or to abandon the LR4 XO.  However, no other XO is obviously better, and the best solutions will likely have to be custom tailored for the room, equipment, and seats.

My own room is kind of a worst-case example of this problem.  The mains and front subs are both located such that the side-wall reflections contribute substantial destructive interference in the XO region frequencies at the seats as well as much of the rest of the room.  The subs also suffer destructive interference from the ceiling.  The MBMs I have are behind the sofa, so reflections affect things less for listeners at the seats, but the ceiling and side-walls still have a negative impact on them, due to their locations.  Anything resembling an ideal LR4 involving the 70-100 Hz range on this system is a complete fantasy.

These problems inspired me to develop a DSP system that allows me to specify filters for each sub unit *and* for each input channel in the XO.  So for example, I might want to utilize my left MBM a lot more than my right MBM at 90 Hz when crossing the front left channel and visa-versa with the front right channel.  The XO can be extensively optimized for each channel and fewer compromises.  Achieving good results for multiple listeners is still very challenging with this approach.  I am able to simulate the raw / unprocessed response at the seats, but the optimization is tedious, difficult, and time consuming.  My approach thus far has been to manually iterate until I give up.  I expect that better tools, supporting FDW and other analysis methods, will help me to get better results.  However, I also plan to add more mid bass sources and am still trying to figure out which locations will be most useful. 

At least for now, I'm doing a lot better than I could before.  Previously, I could make separate filters for each sub but they were shared among the XOs for all the channels.  This helped me to reduce multi-listener response variations in the LFE channel, but the benefit largely fell apart in the XO region and above for mains channels.  I did have EQ for the mains channels downstream from bass management, so what I did was run the subs a bit hot through the crossover and adjust the mains as best as I could for a smooth XO response.  The bass through the XO was pretty good in the MLP,, albeit a bit hot, but it was uneven everywhere else.

Anyway, this is still a work in progress.  I finally finished an overhaul of the surround filters using the methods I used for the front speakers.  This included a new bass optimization as well.  All I've heard so far are some of the Atmos demos and other trailers.  Wow!  There's a lot of bass in the surrounds, and the fact that previous configs neglected to fill the XO hole made this quite apparent.  I look forward to trying some movie scenes tomorrow and am hoping my surrounds have enough headroom.  I was often pushing them to near clipping before this new config, so we'll see.  I may need to figure out how to implement a DSP limiter.  :/