Audiolense solutions for correction of a multi-sub array.

[hulkss]

I am using Audiolense (AL) to correct a large subwoofer array. In this case, the array has seven subwoofers of three totally different cabinet designs aligned across a 22-foot-wide sound stage.
There is a Klipschorn in each of the left and right corners, an Infinite Baffle (IB) sub in the middle under the center speaker, and two Epik Empire subs flanking on each side of center.

There is a total of fourteen 15" drivers in those seven cabinets: four in the IB, two in each Epik, and one in each Khorn. There are 5 separately amplified channels: two Khorns; two Epik pairs; and one IB.

I will compare three Audiolense solutions:

• AL measures each sub channel individually as one sub with five drivers.
• AL measures all five sub channels together as a group.
• AL measures all five sub channels together as a group through filters developed with "Multi-Sub Optimizer" software.

I will keep this post as short as I can. Feel free to post comments or questions and I can provide more information.

Solution 1: Here are the five individual sub measurements in AL. Driver 0 & 4 are the Khorns, 1 & 3 are the Epiks, #2 is my home built IB. None of the subs have internal amps or signal processing of any kind. Oddly, they all have the same output at 38 Hz.

To get AL to generate useable filters for the highly different responses it was essential to select the no bass boost option. I could not assign per driver targets. Here is the simulated corrected performance (frequency only correction, minimum phase target).

Solution 2: Here is the result measuring all subs together as a group:

Solution 3: Multi-Sub Optimizer software develops filter sets with the objective of getting the combined sub response as similar as possible in all listening positions. I measured the five subs in four listening positions with Room Equalizer Wizard and used MSO to create a filter set.

Here is the result measuring all subs together as a group through the MSO filters (the MSO filters are loaded into my MOTU 16a audio interface).

Here is the step response of solution 3, it looked the best to me:

Next I'm going to attempt to measure the three AL filter solutions in the four listening positions. I'll use Room Equalizer Wizard software and measure through the AL convolver. This should reveal if MSO does improve the seat-to-seat response variation.

 

[Omid]

Thank you for all the work you put into this. The information is interesting to a lot of us I think.
Interesting that solutions 2 & 3 get rid of the 35 Hz dip seen in solution 1. Curious to see the step responses for 1 and 2.
I’m impressed that your DIY sub is better behaved than the commercial ones.
Also impressive how good your computer step response looks. Have you had a chance to measure it in real life with REW?

 

[juicehifi]

I think the benefit of MSO etc is not found in the sweet spot, but in rather in the form of more even response across several listening seats. And I have some confidence that a better result can be achieved with all three with a bit of tweaking.

hulkss,
Great work and thanks for sharing.
If you send me the measurements I can see what I can do with all of them and will post the tweaks and the results here.

 

[hulkss]

Thank you for all the work you put into this. The information is interesting to a lot of us I think.
Interesting that solutions 2 & 3 get rid of the 35 Hz dip seen in solution 1. Curious to see the step responses for 1 and 2.
I’m impressed that your DIY sub is better behaved than the commercial ones.
Also impressive how good your computer step response looks. Have you had a chance to measure it in real life with REW?

I believe that Infinite Baffle subs are the best. I just cut a hole in the floor and used the crawl space as the back chamber. You need drivers designed for an IB sub. I used these: Acoustic Elegance Here's a few pics:

Here's the step response from the other two AL corrections:

I'll do the REW measurements of the AL corrections soon. I've measured with REW before and had good correlation with the simulations in the original measurement position. This time I'm going to move to my four seating positions.

 

[hulkss]

I think the benefit of MSO etc is not found in the sweet spot, but in rather in the form of more even response across several listening seats. And I have some confidence that a better result can be achieved with all three with a bit of tweaking.

hulkss,
Great work and thanks for sharing.
If you send me the measurements I can see what I can do with all of them and will post the tweaks and the results here.

Thanks Bernt. Please show us how to get the best from Audiolense. I sent an email, the measurement files will not attach here.

 

[2234rew]

This is amazing @hulkss

And very complicated LOL.

I don't know if it gets more complex than this?

Please keep this thread updated as you continue to improve.

 

[hulkss]

This is amazing @hulkss

And very complicated LOL.

I don't know if it gets more complex than this?

Please keep this thread updated as you continue to improve.

Yes, several subs do much better than one. download PowerPoint presentation: Geddes Multiple Subs

 

[2234rew]

Yes, several subs do much better than one. download PowerPoint presentation: Geddes Multiple Subs

Yes that is known - Toole has discussed it a lot too. Multiple subs for smoothest bass response at any position in the room. No dead spots.

Knowing what is ideal is one thing but proper implementation (like you've done) is another thing. It's not easy.

 

[hulkss]

I just measured the AL filter solutions with REW. I used the JRiver WDM driver through the JRiver convolver. The single channel filter was easiest to set-up this way. I only measured the two combined subs solutions as the five driver solution had issues as shown earlier. Bernt is looking at it.

Here are the frequency responses. The bottom two are seat 3 which is the original measurement position. As you can see Audiolense is the real deal delivering a measured response the matches the simulations in nearly identical fashion. Next up are the AL only filter measurements in seats 1-4. Above that are the AL + MSO measurements (a tiny bit better). At the top are MSO filters only measurements. I did not attempt to get my target response with MSO, the goal was to get all four seats as similar as possible. The four seats are mostly within +/- 3 dB of nominal.

Here is the measured step response in the original measurement position AL + MSO.

Here is a waterfall plot of the response in the original measurement position AL + MSO.

Bernt is looking at the AL filters. Maybe I can do something a little better with MSO.

 

[hulkss]

Well, I could not do much better with AL + MSO, maybe 1 db or so. I limited the MSO correction range to 20-80 Hz.

I'm hoping Bernt can figure out a clever process in AL for a multiple sub correction.

 

[hulkss]

Bernt,

Here's a point that may not be apparent in my system. The Khorns are "Small" speakers in the Set-up. They become part of the Sub Array because I send the Sub Woofer channel signal to them. I mix Left Bass with Sub and Right Bass with Sub in my MOTU interface mixer. It is critical to system performance that the LFE channel, and offloaded bass from any source, plays from all speakers in the sub array.

I attempted to achieve the above by using the speaker type "Full Range and Sub" in AL but I could not route bass to other subs from the "Full Range and Sub" Speaker. Maybe this could be achieved in Audiolense?

I do not re-route (offload) bass from the surround speakers. I help them out with a target response that they can reproduce. This has proven to sound the best in my experience.

 

[juicehifi]

Here are my results. After doing some tweaking I found a procedure that worked equally well for all.

18 dB boos is needed for some of the separate drivers in the 5 drive solution. Otherwise, the same procedure works all over. I see here that I used 3 -3 TTD window, Edit: The default 5-3 produces practically the same result for all three. 5 woofer arranngement, more or less as good for the combined sub and workse for the MSO.

I was struggling for a while, not getting the anticipated results. Then I checked the noise removal.

I added some more space before the leading edge. The blue lines are IR after noise removal. This turned out to be quite important. So remember to check that - especially if you struggle to get the desired result.

 

[juicehifi]

I did two targets: One that stopped around 100 Hz and one that went higher. Here's the MSO and 100 Hz:

I was wrong when I said above that 3-5 works equally well for all. The two above was 3- 3 for TTD window the MSO. And next is with 3-5. It produces some pre-ringing:

The fact that the MSO responds worse to 5-3 compared to 3-3 may indicate that there are som elate room reflactions that are difficult to correct in the time domain. But 3-3 is still quite a substantial time window at low frquencies. However, these simulations, and especially what happens before the leading edge on the step response is important. The pre-ringing we see here may be inaudible in isolation, but perhaps expression of something that is audible in an A/B comparison. So the 3-3 window is a clear winner for the MSO.

 

[juicehifi]

Next out is the "combined" where hulkss has measured all subs as one sub in Audiolense.

These are all very good too. (And you can blame the misalignment between IR and STEP and the target on the programmer, btw)

In the nitpicking department, there step isn't as rounded and smooth for the first wave as the target. There is a small anomaly right before the step comes back to zero. This anomaly is also present in the MSO although to a lesser degree there.

 

[juicehifi]

Now for the 5 independent woofers, configured as a 5 channel sub in Audiolense:

These results are exceptionally good, and clearly better than the ones above. The reason is that in the MSO and the combined solution, impulses have been added without being properly synchronized in the time domain. It adds complexity to the signal, and some of that complexity can be difficult to correct. So based on what I see here, my nod goes to the five driver sub - at least for sweet spot listening. The others are very good too. The MSO may get the upper hand for multi seating, but not always.

And I have more....

 

[2234rew]

Now for the 5 independent woofers, configured as a 5 channel sub in Audiolense:

Hi Bernt, for those of us here learning from this thread, how does the 'cross over config screen' look in Audiolense, with 5 channel sub?

 

[juicehifi]

I started off with a target that went to a few 100 Hz. That turned out to be a slam dunk for the five driver sub.

The MSO is "worst" (still pretty good}, the 5 driver is clearly best, and the combined is a moderate version of the MSO.

Here's the MSO:

Note how the frequency response gets uneven above 100 Hz. I attribute that to drivers that are not very well integrated in the time domaing.

The impulse response (below) is quite messy.

And the step response is rugged, although it has the overall shape pretty good.

Compare that with the 5 driver solution below. Note the early roll-off of frequency is the subwoofer low pass filter that is kicking in. It should have been the same for the other speakers (and I have one more thing to look into). Anyway,, much cleaner frequency response.

And the impulse response is in a completely different league here.

And the step response is clearly much better as well.

This shows that e.g. MSO and other solutions aiming for even response across the room should not be used above a certain frequency. Perhaps it is 100 Hz, perhaps it is even lower. I do belive in the MIMO (multiple speakers serving multiple seats) approach, but I am not convinced that it works better in this situation. The time domain behavior was clearly better for the 5 driver solution also when we low passed at 100 Hz. Furthermore, having 5 subs in the room wil do a lot of good towards an even bass response even though the correction is done in the sweet spot. Having said all that,

I have seen a DBA implementation with Audiolense on the top that produced a step response that was so pure it was surreal. So there is no question that a successful MIMO implementation will improve the low frequency behavior.

Two final figures that I interpret as worrysome: The uncorrected responses of the MSO and the combined (below) These are quite messsy impulses, with much more going on in the time domain than desirable. Sometimes it is unaviodable due to diffucult rooms, but here it is due to sub-optimal driver interaction. Sometimes we can correct this ... and it worked very well below 100 Hz here, but this type of behavior adds complexity to the correction. And sometimes that can lead to an audible difference although the simulations are very good looking. So my nod goes to the 5 driver implementation.

 

[juicehifi]

Hi Bernt, for those of us here learning from this thread, how does the 'cross over config screen' look in Audiolense, with 5 channel sub?

Something like this:

The 80 Hz is the bass offloading frequency imported from the bass management tab.

The sub is a separate speaker in its own right, for LFE. The LFE cutoff is independent of the bass offloading frequency. That's the sub doesn't have a low pass frequency here.

All 5 woofers run in parallel here. We could implement crossovers in the sub,, just like in any other speaker.

 

[hulkss]

Thanks Bernt, I picked some great tips from your post. I really want to use that 5 driver solution and not use any digital signal processing outside of Audiolense.

I loaded up your 5 driver solution and looked at the correction filters. Yikes!! I can't use those. AL is attempting to blast out DEEP bass from the Khorns below the horn cutoff frequency. That's just not going to work.

Then I remembered reading about cascaded sub solutions in Audiolense. Hmmm....I guess that means I might be able to put a high pass filter on those two Khorn channels. Here's the new test set-up:

I had to remeasure as Driver 0 won't accept a HP filter and that was a Khorn. No problem.

Now I loaded your TTD settings and a new target with the bottom end up a few Hz. There we go. I think I can use these filters:

Here's the frequency response hiding behind the target, amazing:

Here's the step response:

Now for a minimum delay low latency correction with min phase high pass filters. Not bad at all.

And here's the minimum delay step response. Still pretty good I think.

Now the only trickery I will need to do with the Audiolense output is to mix the Khorn subwoofer signals (the ones with the 30 Hz High Pass) with the Khorn bass signals (Left and Right Bass). I can not figure out how to do that in my AL setup. Here's the full setup I am using:

 

[juicehifi]

Nice!!!

 

[2234rew]

What interface is that handling 18-channels @hulkss ?

 

[Sonnie Parker]

Remarkable and amazing results... bet it sounds phenomenal.

 

[juicehifi]

I agree. And when you consider the capacity to move air....

 

[hulkss]

What interface is that handling 18-channels @hulkss ?

My MOTU 16a has up 32 input channels from the computer over USB or Thunderbolt and 16 analog outputs. The Khorn Bass and Sub signals get mixed so there are 16 analog outputs. The Motu has a computer in it to do routing and mixing with some DSP too.

It also has analog inputs for measurement but not a mic preamp. Most mic preamps are for music not measurement and won't pass deep bass. The MOTU is good down to about 5 Hz according to tests I've seen. I use an Earthworks M30 mic and preamp (I was lucky to find a demo of their small single channel preamp).

The MOTU uses a routing matrix which I think is great. My set-up now looks like this:

 

[hulkss]

I agree. And when you consider the capacity to move air....

My wife thinks the home theater is primarily used for playing measurement sweeps.