Introduction: DIY Soundbar With Built-in DSP

By
About: mechanical engineer by day, noisy neighbor by night.

Building a modern looking soundbar from 1/2" thick kerf-bent plywood. The soundbar has 2 channels (stereo), 2 amplifiers, 2 tweeters, 2 woofers, and 4 passive radiators to help boost low frequencies in this small cabinet. One of the amplifiers has a built-in programmable Digital Signal Processor (DSP) which I use to create 2-way crossovers, custom EQs, and add dynamic bass boost. The DSP amp uses the ADAU1701 processor which is configurable using Analog Devices SigmaStudio (free software). A separate USBi programmer is needed to download the SigmaStudio program on to the processor. Sure offers a not so stellar one for $20, otherwise a more expensive version from Analog Devices can be used.


Main parts list:

Step 1: Kerf Bending the Enclosure

I wanted a unique enclosure which didn't look "boxy" so I decided to use a kerf bending technique to achieve a seamless smooth edge all around the enclosure. I made several (9 per bend) thin-kerf non-through cuts which terminate about ~2mm away from the surface of the plywood sheet. This yielded a rounded edge with a bend radius of approximately 1". Removing material from one face of the wood, allows the plywood to be easily bent. Care must be taken however since this bend is quite fragile. Kerf bending requires knowing the thickness (kerf) of your blade, the thickness of your material, and the radius desired. By knowing these parameters, you can calculate the amount of material removed(number of cuts), outer and inner arc lengths (cut spacing). To make things easier, kerf bending calculators exist but they have a conservative limit on the bend radius. One example can be found here: https://www.blocklayer.com/kerf-spacingeng.aspx

Step 2: Gluing Together

I created a mixture of ~1:1 saw dust and wood glue and used it to fill the cuts in each bend. I tried applying the glue mixture generously since these bend do not have much material left and the bend is fragile. However, once the glue mixture dries, the bend is quite strong (at least strong enough for a speaker). I also created a half-lap joint which is used to join the top piece to the bottom. You could theoretically have one long seamless piece which would be close to 90" long and difficult to handle. Since the bottom is not visible, I opted to split the enclosure in two pieces and have the joints be on the bottom.

Step 3: Making MDF Front Baffle

I used a plunge router and circle cutting jig to cut out the holes for each woofer and passive radiator. I used a large forstner bit and drill press for the tweeter holes. I also used a round-over bit to smooth out the edges of each hole as well as the outer edge of the baffle. I mounted the tweeters as far apart from each other as possible for better imaging but I am not sure how much of an impact this has.

Step 4: Mounting Speakers and Fabric Wrap

To finish the baffle, I rear-mounted all the woofers, passive radiators, and tweeters using 1/2" wood screws. The drivers came with foam gaskets (shipped loose) which created a nice seal when rear mounting. I also used the hole pattern on each gasket to drill my pilot screw holes -- eliminating guessing. I covered the front of the baffle with fabric (attached with staples) and used an adhesive backed foam strip to create a seal between the front baffle and the enclosure.

Step 5: Rear Baffle + Electronics

The rear baffle has a mitered edge which is used to create a flush airtight seal with the enclosure. I used a chamfer bit and a router table to create the 45 degree chamfer and used the same foam strip for creating the seal. The electronics (2 amplifiers, DC power input jack, stereo input jack, and 2 LEDs) are all mounted in the rear baffle. The electronics are mounted in a sealed cavity in the center of the enclosure which separates the left/right channels.

Step 6: DSP Programming/tuning

Digital Signal Processors (DSPs) are widely used in most modern consumer soundbars. Their biggest advantage is that they accept a digital input and can be used for multi-channel sorround sound. For this project, I used the analog inputs because they are easier to design around. The Sure Electronics Jab3-250 amplifier is equipped with an ADAU1701 processor which has 2 input ADCs (analog-to-digital converters) and 4 output DACs (digital-to-analog converters). I used two output DACs to feed each tweeter and two DACs to feed each woofer. Image of my SigmaStudio graphical program is attached and some of the important blocks used are described below:

Input level adjustment: used to decrease the input volume for each channel. I found that this is a critical step that is required for the Dynamic Bass Boost feature to work (described later).

Parameteric EQ: I used a phone app called "Advanced Spectrum Analyzer" to record a frequency sweep (20Hz - 20kHz) and to roughly measure the frequency response of the speaker without any equalizing. This is not the most accurate approach, however, it is fast and it gives me a good starting point without investing in more accurate tools such as a measurement microphone and soundcard for my laptop. I plan to take better measurements in the future and use additional software such as Room EQ Wizard (https://www.roomeqwizard.com) to help me calculate the right EQ. For now, I created a custom parametric EQ which decreases the volume between 500hz and 4000hz. My ears perceived this frequency range louder than the rest. The speaker sounded better (to me) with the volume in this range decreased. Before and after frequency response curves are attached. These are not a true measurement of the speaker's response and most likely very inaccurate but I chose to include them so that I can highlight just how effective a DSP is at altering sound. In the attached graphs, the orange line represents the recorded peak response and the white line represents real-time level (which can be ignored).

Crossover: I used a 4th order Linkwitz-Riley filter set at 3,000 Hz for the low pass filter on the woofers and high pass filter on the tweeters. One of the huge benefits of a DSP is that it can create complex filters such as this with ease. Making a passive 4th order Linkwitz-Riley crossover would require additional components which could easily add up to the cost of the DSP ($35).

Dynamic Bass Boost: Dynamic Bass Boost block provides boost that varies with input-signal level: lower levels require, and receive, more bass than higher levels. Using a variable-Q filter, this block dynamically adjusts the amount of boost. The input level must be decreased in order for the boost to work. This means that the speaker is no longer as loud, however I believe the trade-off is worth it. At 50W / channel, there's plenty of power.

This is my first project with a DSP and SigmaStudio and I am still learning. I will continue to update this Instructable as I fine-tune the sound. I hope you enjoyed the build!

Audio Contest 2018

Participated in the
Audio Contest 2018