- This is my final electronics build to be featured on the blog. I’m too old and blind to spend more time on such things. The six month project was almost shelved near completion due to design flaws in the PCBs purchased from my usual supplier. These were corrected and Zumdish came to the rescue with a brilliant muting module courtesy of the Celestial Department Store vending machine → and it finally worked nicely! It’s a “jukebox on steroids” and the best soundbar on Planet Industro (and perhaps even Planet Quasti)! Good enough for a final project.
An analogue soundbar with active crossovers, four MOSFET chip amps, concealed subwoofers, and stereo Linkwitz Transforms. No Class-D, IR remote controls nor Bluetooth. Just RCA inputs and analogue circuitry. And it packs an almighty punch quite unbelievable and beyond any commercial soundbar!
Background & rantish waffle
Commercial TV soundbars with fashionable “Bluetooth connectivity”, wireless remote “subs”, IR remote controls etc. cater to a market that craves fashion over substance. That’s alien to me (coz I’m not from around here), and they sound terrible anyway. I just wanted good sound quality without all that nonsense. Okay so they’re cheap. Fair enough, but no matter how inexpensive, boomy ill-defined bass and 200ms of lip-sync lag is unacceptable in anyone’s language. For comparison, an internet ping test from my house shows just 7ms to and from a server 30km away! Or is that lightyears? How can anything be that bad?
Very poor judgement and gluttony for punishment “made me” purchase a couple of those cheap things. One was a TCL to complement a cheap TV of the same brand that I had at the time. The soundbar came with a remote wireless “sub”. The other was a JBL all-in-one. The JBL wasn’t so bad, so it gets used elsewhere. The TCL on the other hand was simply unacceptable with tinny sound and among other issues, a weak sub made of thin flims-board with fake black plastic veneer and so poorly integrated that it reproduced distorted voices meaning it couldn’t be located anywhere in the room remote from the TV rendering its “wireless connectivity” superfluous. It’s been relegated to the junkpile!
The TCL had an abundance of latency even when using the analogue audio input from the TV earphone socket. When connected directly to earphones the TV had no apparent lag, so the soundbar was solely to blame. There was an HDMI “ARC” feature to address the issue, but it wanted another HDMI input on the TV that wasn’t available. Then there’s the optical output from most TVs, but there was still distracting lag. They tell you to make sure all DSP is turned off in the TV settings and that it’s set to PCM mode. Doesn’t help. They tell you to adjust the delay in the TV. Doesn’t help. Just adds more.
Nobody watches broadcast TV or streaming services in this room, so a new Android or other TV wasn’t warranted. A Dell UltraSharp 43-inch 4K computer monitor was a better long-term proposition despite being stupidly big for a computer monitor. I can’t understand who this product was aimed at, but it’s perfect for viewing from a lounge chair some metres away. It displays DVD and Blu-ray content upscaled and rendered using MadVR in the media player software on an HP tower PC downstairs and transmitted upstairs via a 15m DisplayPort cable up through the floor. No need for a so-called “silent” HTPC. An older graphics card (Nvidia Quadro K5200) performs the video upscaling/rendering, and the PC can make all the noise it likes. A wireless keyboard controls the downstairs PC through the floor. All the monitor needs for the purpose of this soundbar is an analogue audio output socket. Unusual for a computer monitor, but a reason for choosing the Dell monitor, is that it has one to provide audio signals from its HDMI, DisplayPort, or USB-C inputs to an external sound system. It isn’t an earphone socket. That said, a Sony TV would work too (with IR remote volume control) as that maker is rare in still providing TVs with an earphone output. For such a TV, this soundbar project would provide perfect lip sync for broadcast or streamed material via that socket. For TVs with an optical output, any cheap D:A converter would work. For my purposes the media player software’s internal volume control is perfect and is adjustable using the wireless keyboard or a control app on the Android phone. The media player identifies the Dell monitor’s audio device via the Nvidia driver.
As there is no such thing as negative audio lag, the old TCL TV required the playback software to be set to apply 200ms of video lag. That was disabled for this new soundbar as there is no perceptible latency whatsoever.
The basic requirement for the soundbar was that it provide good lag-free sound with great bass performance from a single unit. Generally, you can’t get tight bass from ported subs or passive radiator designs and they’re too big for a soundbar anyway. “They” will scream otherwise of course, but I ignore them like those who suggest “open baffle sub” is a thing. Sealed subs are always better, but they can benefit from a Linkwitz Transform for deeper LF extension, so that’s what they got. Then a couple of so-called “full-range” drivers handle the rest. They cross at 220Hz with active LR4 crossovers.
The physical design wanted to be slim, so the bass drivers are inconspicuous and face down into forward-opening exit slots under the enclosure. The slim design meant that a 3-way system was unfeasible as tweeters and mids must always be aligned vertically and not side-by-side as seen in other DIY soundbars on the ‘net. Having a low 220Hz crossover and steep (24dB/octave) slopes, it doesn’t matter that the bass drivers face down. 220Hz has a wavelength of over 1.5 metres. Full-range transducers were then the only sensible option for the forward-facing drivers. Just one per stereo channel.
- 300VA Vigortronix toroidal with dual 25V secondaries (a surprisingly good/quiet transformer) mounted off the aluminium plate for the electronics so as not to induce circulating currents in it, and small enough not to require a soft-starter
- Linear ±35V unregulated PSU for the power amps – just a couple of 22,000μF caps after the bridge rectifier
- Linear ±15V regulated supply for the filter boards taking DC from the main supply
- Stereo/twin Linkwitz Transform PCBs
- Stereo LR4 2-way crossover PCB
- Four channel TDA7293 MOSFET chip amps (bass/treble/stereo)
- Dayton Classic 6.5” 4Ω woofers DCS165-4 (designed for tight enclosures and devoid of aesthetics – good for concealing)
- Dayton Reference 4” 4Ω full-range drivers RS100-4 (look pretty and have a reasonable published high frequency response)
- 24Vdc 120mm cooling fan running quietly on around 12V with a capacitor/resistor kick-starter and HF noise filtering
- High fault current safety ground loop breaker
The MOSFET amps are on the large heat sink with Kapton tape/thermal paste insulators. All 4 outputs share a 4-pole Neutrik Speakon connection through the internal partition to the drivers. The 4 speaker returns share another 4-pole Speakon connection and tie together at the copper plate of the main filter caps.
Unbeknownst to me at the outset, the electronics side of the project was jinxed by a few things. One, and I can’t deny it, is that I thought that I’d get away without speaker muting, but far worse were certain “stability issues” in some of my usual PCB supplier’s designs. The boards were nonetheless corrected by under-PCB hacks. The muting functionality of the TDA7293 power amplifier chips should have been exploited but wasn’t, and there was insufficient space to include it. All of this was corrected but only after their belated discovery and much frustration!
Notes re above pics
Constructing the enclosure was a bit more than I could manage on my own without proper carpentry skills nor tools. Nonetheless, it ran relatively smoothly from conception to completion thanks in part to my faithful lab assistant Hansjürg → now a useful member of society, devoid of the neck brace and no longer reliant on pineal fluid injections for daily function. He made the beautiful front speaker baffles and also helped out with a few tools such as Forstner bits and clamps. A few glitches caused by inevitable inaccuracies in the MDF cuts made by the supplier’s buffoons were also sorted out.
Driver compartments up front were dimensioned to the required volumes determined in WinISD simulations by solving simultaneous equations. Electronics in the back with fan cooling for the power amplifier and regulator heatsinks. Having four 4Ω drivers, things could warm up otherwise, and four TDA7293s together draw around 230mA at idle, so good thermal management was essential. The main enclosure is of “25mm” MDF (which isn’t 25mm thick! It’s 25.4mm but they don’t tell you that) with a Tassie Blackwood veneer from left-overs applied after construction. Underneath “sliders” which define the woofer exit slots are 40mm square Blackbutt off-cuts. The front speaker baffles are from 1 inch marine ply off-cuts with Tassie Oak moulding frames – courtesy Hansjürg.
A lot went into the application of the veneer, but similar has been detailed elsewhere on the blog, so I won’t go into it other than by the pics above.
The black dummy panel across the front is 10mm gloss acrylic with embedded 2mm tipped LEDs to indicate that the regulated +ve and -ve rail voltages are good. These are covered by a 5mm clear glass panel of the same size with bevelled edges. The acrylic is screwed down, but the glass end portions fit into recesses at the inner edges of the front speaker baffles. These clamp down on the glass as the speaker bolts are tightened into concealed tee-nuts. The baffles then hide the screws in the acrylic. 15mm Tassie Oak quad mouldings trim the glass at the top and bottom. 1mm x 15mm adhesive neoprene tape fills gaps and provides rattle-proofing.
Four “Mickey Mouse ears” cut into the MDF around each front speaker cut-out enable better interaction of the driver piston with the sealed volume of each chamber. The MDF is an inch thick, and the plywood baffles are roughly the same thickness. That’s a 2 inch thick baffle! Ordinarily the driver’s magnet would sit well inside an enclosure in free space away from a much thinner baffle (typically 18mm), but here the magnet would almost block a standard round cut-out in the MDF leaving just a narrow annulus. It had to be enlarged. With the four 25mm dia. holes pre-cut using a spade bit, the main cut-out followed with a hole-saw. The area of the annulus plus Mickey Mouse ears is around 50% of the driver’s piston area. Good enough for 220Hz up.
The electronics are on an aluminium plate to enable removal for servicing. It is bolted down to tee-nuts under the floor.
A steel access panel provides access for adjustment of the multi-turn output level trimmers on the active crossover PCB and to some of the aluminium plate’s hold-down bolts and Neutrik Speakon plugs. The steel plate is held down by super magnets bolted in and concealed under the veneer. A magnetic lifting tool was made to raise the panel.
The +ve outputs of the 4 power amps share a 4-conductor cable and a 4-pole Speakon connection to the driver chambers. A second 4-pole Speakon connects the -ve speaker returns directly to the centre of the copper plate tying the main PSU caps at the zero volt line.
Power-up tests resulted in almighty “musket shots” from a sacrificial test speaker. Also “radio” interference from the test speaker followed. The radio interference was a kind of swishing/swirling noise. The Linkwitz Transform boards were isolated as the culprits for the latter, and these were examined with the oscilloscope and corrected with under-PCB hacks as noted above as well as the installation of extra 100nF ceramic bypass caps from each opamp supply pin to ground which were strangely omitted from the boards. The musket fire was reduced in amplitude but still, the system required some kind of turn-on delay muting to be added.
I had consciously chosen not to bother with the usual muting and speaker protection regimes deployed in previous builds as this is “only a soundbar” with inexpensive drivers right? A little turn-on/off thump never bothered anyone. Huh! Also, after powering down, the NE5532 opamps demonstrated unsophisticated flatulence after a 4 or 5 seconds as the rails collapsed past their specified minimum supply rating. I hadn’t the space to install a secondary low voltage transformer to power a regulated supply with an auxiliary relay output, nor provide any loss of AC detection for rapid muting on removal of power, nor was there sufficient space for 4 high-current muting relays to protect inexpensive drivers costing less than such protection! Instead, a simple ±15V regulated supply for the opamp circuits takes its input as DC by dropping the ±35V main supply rails to around ±25V using 150Ω series resistors ahead of the pre-regulator smoothing caps as determined by the load current. The opamp circuits and LEDs require 110mA. Something else was needed for muting.
Celestial Department Store to the rescue!
My good friend Zumdish took one look at this project and described it as “shoddy, shoddy and primitive!” and that he’d melted down such items over a century ago. By sheer coincidence however, his salvage department had one NOS muting board left over (as once installed to ensure dumbness in android security guards) and it was exactly the same size as the active crossover PCB which enabled tidy stacking! 😊
The “CDS Deluxe DC Muter Model 77B” has a 555 timer, 3.3V Zeners, function-indicator LEDs and a bunch of other components. It takes the +ve regulated DC rail to activate a pair of on-board miniature DPDT communications relays to un-ground the four outputs of the active crossover about 2 seconds after powering up. They ground the inputs to the power amps in their default mute state. Turn-on noises (gunshots) are suppressed.
A comparator releases an up-front control relay as soon as the +ve DC rail drops to around 12V after powering down. This grounds the outputs instantaneously and well before the NE5532s let off their obnoxious noises (at around 5V). The -ve rail is loaded by an array of onboard resistors for symmetry. The muting board requires 60mA from each rail, so the series resistors up front of the regulated power supply were changed from 150Ω to 75Ω (by adding parallel 150Ω resistors under the PCB) so that the regulators see about 21V with a load of around 170mA. Crude but effective.
There’s a PDF here: CDS Deluxe 77B. Order yours today and don’t miss out! 🙂
The cooling fan was an essential part of the design, but it had to run quietly. It’s a 24V DC brushless fan taking power from the +ve 35V rail of the main PSU. Series resistors drop the fan voltage to around 12V for slow spinning, but the fan needed a kick start. That came in the form of a 470μF capacitor across one of the resistors to short it out for a few hundred milliseconds on start-up. Then there was a minor electrical noise of the fan heard through the high efficiency test speaker. Disconnecting the fan supply leads from a terminal block shared by the power amp modules, and connection directly back to the main +ve filter cap for proper star grounding pretty much eliminated the noise, but I took it a step further with a filtering circuit comprising series inductors and cross-rail capacitors. Anyway, there’s no audible noise of the fan from the speakers.
Installed and working
The soundbar performs much better than anticipated. The sound is probably superior to any commercial soundbar. The bass performance is pretty staggering considering the purpose of the thing and its meek appearance. I was quite shocked that the assumed piss-ant TDA7293 chip amps could drive the tiny woofers so hard that the floor of the listening room shook under my feet. The upper range has a nice sparkle to it. All that said, it’s not hi-fi! More like a jukebox on steroids.
I’d post the acoustic measurements, but it seems people have neither interest in, nor understanding of such things. Suffice to say that the frequency response – especially from 3kHz to 15kHz – is almost dead flat. Surprising for such inexpensive “full-range” drivers. Only mild EQ was required outside what was applied to compensate for room effects on the bass. There is typically around 0.5% THD even down toward 30Hz where SPL drops off rapidly and THD rises only to around 2.5% at the reference level of around 80dB.
That’s it. Over and out. 🙂