• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DIY 2.2 living room project

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,645
Likes
1,914
Location
Budapest
One of my friends came over a few weeks ago to listen to my new DOS sub (that was placed in my home office system for testing purposes before being moved to my living room)
He loved it so much that he decided to build the same for his living room :)

Before venturing into this project we used one of his existing DIY DOS subwoofers (from his home theater system) to measure the best position for the future sub in the living room
Quickly we noticed that one sub will not be enough to optimize the full bass range so we identified two positions in the room that when combined shall provide pretty good results

This is his living room:
IMG-2ef41da4af4147c14cadae8e2bb34e43-V.jpg


His MLP is in the middle of the couch

One of the new subs will be placed behind the couch, right behind his MLP (as shown in the below picture that was taken while doing the measurements in search for the best positions)

IMG-40a64153ebc100aaead8cca3d0a2b617-V.jpg


The front sub will be placed next to his front right speaker (where you see the cardboard box on the floor)

IMG-d59fb94133a7155d36ad694650d03a58-V.jpg


This is the response we got in the MLP when placing the above test sub behind the couch:

1729678721479.png


And this is the response when placing the test sub in the front:

1729679349393.png



This is the combined response:

1729679403454.png


And this is the predicted EQd result:

1729679464835.png


A significant amount of boost will be needed in the low end but it shall be fine since he listens at low to moderate SPL (the amps will be able to cope with that)
I am attaching the mdat file for those who are interested in all the graphs

So after he saw the graphs he got really excited (quite understandably) and started the build immediately :)

This is the design using the same Lavoce NBASS08-20 drivers (two per cabinet) that I used in my project

1729679765797.png


Each cabinet will be powered by a Fosi V3 Mono amp
When driven with 240W of power they shall each produce 95dB at 24Hz at 1 meter (=101dB for the two cabinets)

1729681498072.png


All this at 86% of the linear xmax

1729681560702.png


Here is how the build is progressing:

IMG-72932360230260822e93cb7eaea2c130-V.jpg


IMG-6a0c22ae73d876001c981c981f154bcf-V.jpg


IMG-bfdcbc79823bacf41afe70ed4ea4e947-V.jpg


In the last picture you can already see the 'beauty front' from Ikea :)

The drivers shall arrive in one month's time, I will post an update then
Stay tuned!
 

Attachments

  • living room two subs.zip
    2 MB · Views: 32
Forgot to mention that a Focusrite Scarlett 4i4 4th Gen. audio interface will be used as a multichannel DAC (to be able to manage all four channels separately)
According to the manufacturer's specs it shall have a SINAD of 112dB - which is OK

IMG-34017efd168a2e1247fb21083341d096-V.jpg


Jriver will be used as the DSP with Dirac Live as the DRC
The subs will be integrated 'manually' so Dirac will 'see' the system as stereo only (= no DLBC will be used)
Crossovers will be linear phase, probably at 80Hz with a slope of 24dB/oct. using thEQorange VST plugins
 
Just an update to inform those that consider buying the new 4th Gen. Scarlett interfaces for multichannel/multiway usage
I could not figure out how to set the interface to treat the four channels totally separately but after contacting Focusrite's support it got resolved

You need to set up Control like this:

Configure MixA to route to Output 1-2

1729855307540.png



and configure MixB to route to Output 3-4

1729855359136.png


These mixes will run in parallel so this way all four channels will be active separately; i.e. one can now configure volume, delay, EQ, etc. per channel in Jriver (or whichever player/DAW)

FYI @MusicIsADream I think you were looking for something like this on a Mac, as you see Focusrite's screenshots are from a Mac so this shall work there too
 
Hi ppataki. I like your builds, and I'm jealous of the energy you have for pumping them out, but man, you really gotta step up your room/speaker integration skills! Because you are both missing out on performance, whilst not effectively negating the room problems. I'll give some info below, but as you will see it is a time consuming process. If you have DLBC and individual output channels, then the quick and easy solution is to let it control everything. Let it handle the delays, the filters, and the rest of the integration, and then you don't have to get bogged down in it yourself.

Anyway, I downloaded the mdat from the first post. Straight away I can see there is no timing reference or offset information, and the impulses are just at IR-max = T0. If you are manually aligning speakers to a listening position, you need to know the timing information. And it's not just a delay because of distance, but high/low pass filters and ported-vs-sealed also comes in to play for defining group delay, and thus arrival time. (Now it is possible that you already set some delays and "baked" it in to the measurements, but I doubt it.)

Why does it matter? Well, for your combined response you went and summed the 2x IR-max aligned impulses together, but you don't know if IR-max actually arrives at the listening position at the same time. Even if they do happen to arrive at the same time (perhaps because you set the delays already, or perhaps by chance), aligning at IR-max is not necessarily the best option. In fact, you can see it is not the best option just by looking at the raw impulses on the overlay screen;
impulses.png

When working on impulse alignment, you don't want to be using the basic A+B tool, but the Alignment Tool on Impulse mode. Here is a screenshot of the tool showing the impulses filtered at 60Hz, with no delay or gain set. You can see that the summed freq response (black line) matches your basic A+B combined response (highlighted underneath). You will also see a line is a dashed black line, and that is the theoretical best summation for these frequency responses if you do not consider phase issues. So the closer you can get to that dashed line, the better. And as you can see, there's a lot of gaps (gaps = sacrificed performance). Also pay attention to the filtered impulse at the bottom. Even though they are aligned IR-max, they start off similar, but green "stretches" (has phase shift), and ends up offset (cancelling) red. That's all room affect on the impulse.
no_alignment.png

Now as an example, here I randomly picked 60Hz and let REW work out the alignment. You can see straight away that the impulses are far better aligned time wise, and that a bunch of SPL has been gained from 45-75Hz. You can see that a ~15ms delay and a signal inversion is required to get that alignment.
auto-align-60hz.png

But in that example, some areas like 88Hz got worse. I can investigate why by moving the cursor to that frequency, and filtering the IR's at that point (**filtering, not aligning, at that point).
60hz_auto_align-IR_at_88Hz.png

Now you can see why SPL was lost at this point, and it's because the impulses are phase-shifted. Green impulse is the culprit, because whilst it starts off okay, eventually a bit of ringing (from a room mode) comes in and screws up it's impulse. That's also why green has lower SPL vs red at this frequency - the ringing is causing it's own cancellation (ie, it's a weak null). And at this point is where the magic can happen, because you don't have to live with that combined response, and you can instead EQ each subwoofer individually. I don't have time to explore the various options for you, so I just threw a notch filter on green at 88Hz, then aligned with exactly the same specs as found before. And voilà, the combined response is improved!
notch+align-60hz.png

But there's no need to stop there, as the dashed line is indicating that there is still more to gain. More EQ to target the problems of each specific sub is the idea, but note that EQ'ing to get the best response possible for an individual sub is not necessarily the best for "the subs" (the alignement). Because for example, the ringing of one could be used to cancel the ringing of another, like a poor mans Dirac ART! So here I quickly explore such a process by manually sliding the offset sliders until I found something I liked, and which looks a bit "different".
manual_align.png

With that alignment, I have secured nearly all of the performance possible above 50Hz, but it kind of looks worse below that, especially at 40Hz. But don't let 40Hz fool you, because that's not "clean" bass - that's a room mode causing a big boost (besides the SPL peak, you can tell by the christmas-tree shaped IR). So what I've done with this particular alignment is partially offset those IR's and put them in to cancellation, reducing the peak (the boom) by reducing the ringing. You can see this in the spectrograms that follow, because the red are at 40Hz is reduced in length/time. And overall, I think it actually leaves a nice looking rolloff from 50Hz down to 20Hz, which looks like decent room gain for 8" subwoofer (and is probably quite agreeable to adding a "bass boost" shelf).
combined vs aligned_v2.png

Here's an SPL graph comparing your summation versus my evolving alignments. (BTW, I like to work with var smoothing because it gives "clarity" where you need it, ie more down low, less up high).
compare_sums.png

So yer, ther is a lot of possibilities when integrating at this level. A couple of other things to look at All-Pass filters for tweaking certain phase issues, and another option is strong PK filters with high Q-ratios (eg -10dB and Q=10), as these cause ringing that can work to cancel a room modes ringing. REW's EQ tool can actually work this out for you, but you have to drill down through it's menus and configure the room settings, and let it analyse the modes. Once it has, it can then make better EQ choices by first targeting the modes, and then fixing the result. This is opposed to just "EQ to target" like it seems you did, as that often just uses a bunch of wide/low-Q filters to make the freq response flat, whilst leaving all the phase/timing related problems unsolved.

Anyway, that's the end of my TED talk. Hopefully this info can inspire you further :D I was acutally going to post something like this in your dual-8" office subwoofer thread, but I didn't get around to it. The same jist applies though. But actually, I would highly recommend you look in to and try out a cardioid arrangement in either of these rooms (Cardioid in this context as in integrating 2 nearby subwoofers to work in cardioid configuration). The benefit of going cardioid is that the setup (in simple terms) outputs somewhat directional bass at a target frequency, versus 360° bass. This can be used to send directional bass down the rooms main mode, and because there is less "back wave", there can be way less ringing.
 
Last edited:
Thanks @neRok for the detailed explanation, appreciated!

Last time when I tried DLBC it sounded absolutely horrible compared to doing the integration manually - having said that I might give it a try again with my new UMIK-2 mic
(I am waiting for my new amp to arrive in my living room so I can finish my 4.2 system - which is different from the system discussed in this thread; this one belongs to one of my friends)

In the mdat file the sum of the two subs was just created by the A+B operation in REW, just to have a rough idea how they will work together
Both the subs and the front speakers will all be sealed and the crossover will be linear phase so I guess those will not really have an impact
Once the project is done (hopefully sometime in December) I will post all the measurements, including the sub integration in an mdat and then you will be able to scrutinize it if there is any more room for improvement
 
A few updates in the meantime:

viber_image_2024-11-15_19-41-58-867.jpg


viber_image_2024-11-15_21-53-52-888.jpg


viber_image_2024-11-17_14-44-15-103.jpg


viber_image_2024-11-17_19-41-59-236.jpg


Build is almost finished!

The setup and the calibration will take place hopefully in 3-4 weeks' time
I will post all measurements accordingly
 
The build has been finished completely:

IMG-d656cc9e2c4e299bba874916cade6680-V.jpg


IMG-29fb3ac4acb6c4a19aeec6fd96bad94e-V.jpg


The second sub looks identical and it is behind the couch

The two Fosi V3 Monos are also in place with the Scarlett 4i4 4th gen DAC, all the cabling is done too
Calibration and setup will take place on 7th December by me, I will post more pictures then and all the measurements
 
As a life long DIY speaker builder I really enjoyed checking out your approach. Using deconstructed IKEA products as your lumber yard is quite an original approach. The results speak for themselves. Nice work!
 
Those look really great, nice work! Definitely fits the look of the room as well as, or better than, anything you might buy off the shelf. One of the best reasons to DIY these days.
 
Yesterday I went to my friend's place and we calibrated the system
I must say upfront that the end result is simply fantastic - you cannot locate the subs at all by ear, the lows are super-fast and precise. All three of us (and the wife and kids) were all totally blown away.

Let's start with the measurements:

L+R before/after a manual pre-correction (linear phase low shelf) to ensure relatively OK response in the lows too:

1733676558405.png


Front sub and rear sub:

1733676644466.png


And now the integrated response after delay settings and crossover (80Hz, 24dB/oct, linear phase)
Absolutely no filters were applied to the subs
To be honest I did not play around with the Alignment Tool, I just applied the measured delay values since we were pretty happy with these results - but I am sure it could be further optimized

1733676995777.png


Let's see what Dirac did with it (1-point optimization in the MLP)

1733677129815.png


And the full 20-20kHz view:

1733676890955.png


Distortion is fine:

1733677207198.png


The room has enormous reverb unfortunately, which is visible on all temporal curves

1733677399916.png


Group delay is a bit higher than I would have anticipated but I guess it is still way lower than what for example DLBC would have done

1733677459151.png


Wavelet shows <20ms of peak energy delay which is OK

1733677505836.png


So again, as mentioned above we were really very happy with the results :)

And finally some more pictures, a video and the mdat file for your scrutiny:

1733677701810.png


1733677738388.png



 

Attachments

  • 1733676743138.png
    1733676743138.png
    254.5 KB · Views: 9
Group delay is a bit higher than I would have anticipated but I guess it is still way lower than what for example DLBC would have done

Those jagged peaks are just from normal room reflections -- but importantly the ~20 ms peak energy sub frequency range is already quite good. 30-40 ms is borderline still okay and above 50 ms should be fixed/avoided if at all possible.
 
Back
Top Bottom