A pair of 500 Watt (4Ω load) Class-B subwoofer plate amplifiers each have ten bi-polar output devices, linear PSUs, 500VA transformers, inrush current limiters, speaker protection and muting modules, over-sized heatsinks, thermal protection and quiet fan cooling. They’re engineered to my standards, not theirs and they’re “Super”! 😀
Most ready-made subwoofer plate amplifiers sold in the past ten years or so fail shortly after the warranty period, or sooner. Maybe it’s the unreliable lead-free solder used for RoHS compliance, something more nefarious, or generally crummy design. I don’t know. I do know that they usually have vast arrays of unnecessary connectors like XLR and speaker level inputs, satellite and daisy-chain outputs, dials and switches galore – all to impress the wide-eyed, but skimp on fundamental things that matter like transformer VA, the number of output devices, plate thickness, heatsinks, soft starters, speaker protection and thump suppression. In the case of Class-D models (at least), the use of SMDs renders them virtually irreparable when they fail anyway. And switching power supplies are just bargain-basement noise generators in my book and must be avoided.
Here I’ve built a pair of simple (or maybe not so) and robust Class-B subwoofer plate amplifiers (with carefully executed through-hole Sn-Pb solder joints) to replace very disappointing Hypex (or is it HiVex?) units. Built with plates of identical footprint to retro-fit into existing DIY speakers, these amps are jam-packed straight-through old school powerhouses without so much as a volume control and have nothing to access externally but a switched power entry receptacle and an RCA input socket. The amps can deliver about 250W RMS into 8Ω, and into the load of the actual speakers, they can deliver about 500W RMS. None of those silly exaggerated advertising figures here! And the amplifiers won’t end up as land-fill. Not in my lifetime anyway.
This is not a conventional subwoofer plate amplifier application. The amplifiers are not intended for remotely-located “subs”. They’re for the parallel 8Ω 12 inch Morel woofers in the sides of these 3-way actively crossed DIY sealed speakers, built back in 2011:
Those woofers are something a bit special with voice coils that surround the magnets for efficient cooling.
The amps replace failed Hypex plate amps, so needed to fit into tight amplifier sub-enclosures near the base at the back, which were sized for the original amps.
Just 240 × 240 × 80mm deep.
Had it not been necessary to squeeze everything into such a tight space, the new amps would have taken on a very different form!
Choice of amplifier type
The best subwoofer amplifier in my experience is this two-channel P68. It’s straight from Planet Industro, so is of no-nonsense design with better than adequate specifications (exceeding human hearing thresholds), is totally reliable and kicks some seriously fat butt. It’s better than a big Krell amplifier in its intended application. So it was decided to build plates using the same amplifier modules. The output devices are the most rugged and linear bipolar transistors available from ON Semi – MJL4281A (NPN) and MJL4302A (PNP). BJT devices are more suited to high current subwoofer applications than MOSFETs. And switching power supplies are absolutely out of the question, so conventional and necessarily massive linear PSUs exacerbated the “squeeze” problem. At least the transformers could be down-scaled from the 1kVA unit seen there to 500VA for single channels.
No more Class-D for me!
Before ordering that through the Celestial Department Store’s vending machine → however, I had a number of Hypex plate amps, which had all become a right PITA. The DIY Class-D amp with very early non RoHS modules has been fine so far, but IMO their plate amps are poorly designed, unreliable and inconsistent junk piles – from straight out of the box DOA returns and early failures to production variants in the same model with unadvertised and unwanted voltage gain changes. One amp had 6dB more gain than another! One of their most annoying features was a flimsy input filter board with “go open” pots that did nothing useful and could not be bypassed. One such stupid feature was a phase manipulation dial that worked with a phase inversion switch. The dial was marked “0°/180° •• 180°/360°” which is absolutely meaningless without specifying the frequency at which these angles occur, and the switch is marked “0°/180°”. It’s one thing to provide stupid and unwanted filtering, but it must have a bypass for people with active crossovers! And the plates were inconspicuously leaky so couldn’t be mounted to woofer enclosures without air chuffing through the apertures around pot shafts, RCA sockets etc. A separate enclosure is essential but they didn’t tell you that! And despite Class-D efficiency and undersized transformers that ought to be fine, they never produced anything close the claimed output power. One of the supposedly 800 Watt (yeah right!) amps started to produce random thumping sounds from the connected speakers even without an input signal. Two service enquiries with the manufacturer went unanswered. IMO the company stinks, so I have no interest in even looking at their latest offerings. They get the deserved wrap right here. So those PoS have gone to the cornfield in favour of “real” amplifiers that – given the patience – any half-arsed DIY type could build.
Another broken and unserviceable PoS back in its coffin
Not even good for parts. Even the transformer is unuseable.
Aside (up-front band filtering)
The amplifiers were not to include any of the unwanted filtering that commercial plate amps often provide. They serve as power amplifiers and therefore have only a basic and innocuous low order high-pass filter at the input and no volume control – just like the other power amps in the active three-way system.
Another of the coffin-dwellers’ unwanted “features” was a 24dB/octave infrasonic filter cornering at 12Hz that could not be bypassed. Commercial plate amps can benefit from an option for some kind of over-excursion limiter for home theatre use, but the filter in thw Hypex was a big compromise on fidelity because it caused phase shifts right up into the upper end of the subwoofer’s range to screw with the midrange crossover alignment. The best I could do at the time was to build the up-front active crossover with that in mind – i.e. without some of the normal inter-stage DC-blocking caps.
So for the new amps the active crossover needed modification to provide additional high-pass inter-stage filters each with corners aligned between around 4 and 6Hz.
The simulation – simplified with opamp buffers representing the various stages between the HP filters:
It produces an overall response being a 24dB/octave roll-off that’s about 1.5dB down at 20Hz and about 5dB down at 10Hz:
While not of the same sharp Q as a dedicated rumble filter such as ESP P99, the above nonetheless protects the drivers from over-excursion, reduces the likelihood of amplifier clipping on extreme low frequencies and mitigates false tripping of the speaker protection module (see below), yet still allows for very deep bass (less than 3dB down at 16Hz for sub-contra C organ pedal) that the room will more than compensate for.
The active crossover crosses the signal to the midrange with an LR4 set at 150Hz, so the plate amps basically receive a 20-150Hz band with 4th order slopes at both ends.
Active crossover modified by installing 8 caps with ones of correct value per the above simulation (0.66μF is from paralleled pairs of 0.33μF caps):
In my set-up there are additional low order HP filter stages in a subwoofer constant Q graphic equaliser and a DBX 120X-DS sub-synth with very low corners. These are ahead of the preamp and should have little influence on the predicted response.
The point of all the above is that infrasonic filtering must be applied only ahead of the active crossover, so that its phase response – which extends right up and across the bass/mid crossover point (it’s 3° at 150Hz) – does not disrupt the LR4 crossover’s coherent phase alignment. Put another way – the same 3° of phase error at 150Hz also applies to the midrange driver’s signal, so the transducers align perfectly.
Power supply boards (beware – more ranting)
Chinese insults from eBay labelled “Power Provide BY:Bobhifi” [sic]. Well Bobhifi (whoever you are) your layout is about as sensible as your English. Issues included a module bridge rectifier footprint having a mislocated through-slot requiring a leg to be bent. It also very stupidly provided for an additional parallel rectifier in a SIP package (with through-holes correctly spaced) but what kind of spastic does that Bobhifi? WTF were you thinking? Anyway it was the smallest PCB I could find with provision for 4 electrolytic caps, so bending a leg it was.
The board had other evidence of gross incompetence in unnecessarily narrow and unsafe track clearances of a mere 0.5mm. Why? Even with solder-resist it’s too fine. There was plenty of space to lay the board out sensibly. Those clearances might just be OK for say ±20V rails, but no way for ±70V, so a little gouging was required.
0.5mm from +ve DC to 0V – increased to around 2mm:
0.5mm from AC to DC rail increased to about 2mm:
Notice also the ridiculously close proximity of that AC trace (top left) to the stand-off hole. The solder-resist is not enough. Anyway I used a plastic stand-off there.
Rail tracks cut for a couple of 0.1Ω 5W resistors to form π filters with the main caps:
All cuts enamelled-over to reinstate the solder-resist. Also several of the tracks were unnecessarily narrow for high-current duty, so additional heavy gauge Teflon-sheathed wires were added. Caps are the highest μF that I could source in 30mm Ø to suit the PCB at 6,800μF/80V (13,600μF per rail).
Finished and safe boards (no one will ever know). 🙂
There’s a processor heatsink thermal epoxied to each bridge rectifier.
More incompetence in the provision of only a single zero Volt connection at the output end. Useless for mono, let alone stereo amplifiers. Where do you expect anyone to connect the speaker return Bobhifi? I suggest you get another job – perhaps as a janitor. Anyway I jimmied something under each board for a second cable connection.
I almost gave up waiting for the local manufacturer Harbuch to provide these stock transformers and was contemplating abandoning Aussie made for cheaper European transformers from RS or E14 with 230V primaries, but I’m glad I was patient. These are very nicely made and have proper 240V primary windings, so they run quietly and the 50V secondaries will be slightly under voltage on a 230V supply. That was part of the plan – a little peace of mind for the 80Vdc rated filter caps which will now see around 67Vdc instead of 70 at idle:
These 500VA transformers have a very substantial core and weigh more than the equivalent-rated transformers from the usual suppliers. The very small mounting hole means a bigger core for reduced magnetostriction noise. They’re actually heavier than most complete commercial plate amps!
Seems I’ve assembled a hundred of these over the years. They’re still the best/most compact soft starter available. No incompetence here – ESP P39. I used compact 150Ω 7W high surge current wirewound resistors for the ballast and a 100μF cap at C2 which increases the second relay activation to about 1 second. Previous builds had huge banks of capacitors and much bigger transformers up to lkVA and slower timing was used. This is a mono amp with just a 500VA transformer and a capacitor bank that isn’t OTT, so 100uF is about right. This also ensures that activation beats the speaker protection relay (below).
9V Auxiliary transformers
These power the soft starters which rectify to 12Vdc for the speaker protection modules (below) and the cooling fans. The space is so tight that the connecting tabs had to be bent to clear the back of the fan inlet manifold (below).
Speaker protection modules
The usual P33 with component substitutions that provide a 2 second delay before the speakers are connected on start-up and a Zener diode is added to speed up the relay release should a fault be detected. These are designed for stereo amps, but a resistor and a capacitor are omitted for mono. They are set for “full range” use – i.e. only very low frequencies with some amplitude (or DC) will trip them into action. The 15A relays are mounted off-board. One corner of each PCB had to be rounded to protect the power transformer from gouging:
Two ESP P68 modules were built to specification except I used resistors in most locations of at least double the recommended power rating. I seem to recall that 0.6W parts at R10 and R11 for example were marginal with 70V rails, so 2W metal film parts were installed. The smaller red ones are 1W 5% metal film parts that measured within 1% of spec anyway.
Copper was used instead of aluminium for the driver transistor/diode heatsink mount because it’s a better thermal conductor:
There are sil-pads and plastic washers on the driver transistors and the two diodes are thermal epoxied to the copper plate. There’s a graphite pad sandwiched between the aluminium heatsink extrusion and the copper:
The modules have ordinary single-ended input connections for the RCA sockets because balanced inputs (XLR) are not warranted in any home audio despite what they say.
Extra output device boards
The speakers each have a pair of high power-capable Morel 12″ 8Ω woofers connected in parallel for nominal 4Ω loads. I often use sub-synth, and the active crossover runs a Linkwitz Transform in each channel that calls for around 17dB of LF boost, so the amplifiers needed to be extremely rugged. The additional 4 output devices and high rail voltages were therefore mandatory.
These auxiliary boards are off-cuts or “half boards” of main P68 PCBs and are intended to be connected in-line with the main board, but there was insufficient space in this build, so they had to be turned sideways. Solid core jumpers through Teflon tubes join the boards:
They tested OK on an old wooden chopping board with 315mA rail fuses and a ±12V mini bench supply:
Well the green LEDs illuminated dimly and there was a DC offset of around 110mV at the output of both assemblies. Current drawn was less than 1mA. The quiescent current should increase when ±70V is applied assuming nothing blows up.
Insulators are the usual 25μm Kapton tape:
These were cut from 5mm thick aluminium stock. Edges chamfered, holes drilled to markings made through the PCBs. Larger holes were step-drilled and rectangles first drilled then filed. Lots and lots of countersinking too. Etch primer and satin black enamel. Heat conduction areas masked off:
↑ By far the most tedious part of the project! One pic shows them baking in the sun after enamelling, and another shows a handle bolted down. These are 40mm high – the same height as the heatsinks. They assisted in the build to keep the amps level on the workbench and provide something other than the heatsink to grab while installing the amps into the speakers.
These 200×160×40mm extruded aluminium sinks had holes tapped to align with those drilled through the plate – ten M3 for the output transistors and a couple of M4 to stabilise the other side of the sink to the plate.
I must thank my good friend Dr Adrian (Head of Celestial Department Store Service Dept.) who kindly did the thread tapping for me. I have no such skill! 🙁
Both mating surfaces were honed flat with an orbital sander and Scotch-Brited before applying high quality thermal compound.
Safety Ground Loop Breakers
Space was at an absolute premium, so I used compact 30A SIP rectifiers that were initially intended for the PSU but not used:
These connect between the floating star grounding bolt and the mains-Earthed plate and eliminate speaker hum from circulating currents around the interconnect cable and the mains lead.
A bit hard to see in the photos, but there are black fibreglass insulator boards installed under the high voltage PCBs for safety.
As noted above, the amplifiers are designed to be installed in loudspeakers that have a separate sealed amplifier enclosure. The fan cooling would not work as intended if installed directly into a speaker enclosure (especially one with a port). This is what Hypex should have stated, but it would have cost them sales if buyers knew what they were in for. In my case it is all to do with the cooling regime.
This 60×60×15mm 12Vdc fan while advertised as “quiet” at Element14 is a screamer when operated at the nominal voltage.
Only a limited air flow is required to compensate for the lack of passive ventilation slots, so I tried slowing the fan with series resistors until finding the ideal compromise between airflow and quietness. That was with a 150Ω resistor, but the fan wouldn’t start with just 5V across it. A flick of the blades would get it going below its starting voltage threshold, but that’s no good. The answer was to place a capacitor across the resistor so that the fan sees the full 12V for a while (guess around 100ms or so) at start-up. 220μF was enough, but for a safe margin I ended up using 470μF. Electrical switching noise from the fan via its power supply leads is isolated from the amplifier by its separate power supply.
Vdrop resistors and caps on bits of perf-board:
The fan is mounted to an inlet manifold to draw air in from outside of the amplifier. It’s made of fibreglass boards and 6mm square aluminium and PVC bar:
A couple of brass M3 tapped 6mm blocks flank the inlet:
But once the fan was mounted to the manifold it wouldn’t draw much air at all. The manifold is narrow by necessity to fit within the amplifier and leaves just 6 or 7mm behind the fan blades. Rotation of the blades created turbulence within the tight inlet space and as such the fan’s efficiency was reduced significantly. The answer was to ensure laminar flow to the blades by fitting perforated aluminium screens directly behind them to work as airflow straighteners:
There are off-the-shelf PVC filters like these that might have done the trick too:
… but they’re flexible and might have bent into the blades and would clog up with dust pretty quickly anyway, so here’s how one of the aluminium screens looks installed:
And they work really well. Many smoke trail tests were done using incense sticks! 🙂
The assembly bolts down over these inlet holes in the plate:
There’s an outside cover plate cut from aluminium sheet, with spacers.
The bolts go through to the brass blocks inside.
And it hides the fact that I drilled the holes slightly out of line. 😀
There are black spring washers there above the fibrous washers.
So that’s the air inlet. The fan draws air into the amplifier enclosure through those five 6mm ∅ holes to cool the electronics. It blows air across the PSU’s bridge rectifier heatsink to pressurise the enclosure very slightly and provides general circulation.
The air escapes through the plate via ten 3mm ∅ holes to be deflected by a re-purposed kitchen drawer handle to cool the external heatsink.
The exit holes are underneath the input end of the main P68 board. The board has about 5mm (thickness of the TO-264 transistor packages) clearance underneath for good airflow. The airflow path will be like this:
The total cross sectional area of the five inlet holes is about 141mm², whereas that of the exit holes is only half that at about 71mm². Taking thermal expansion into account, the exit velocity is more than double the inlet velocity for good turbulence where it’s needed – between the fins of the main heatsink.
Inspiration for the cooling regime came from a vintage Cerwin-Vega professional amplifier, although my implementation is quite different.
Additional thermal protection
A 60° NC thermostat is mounted to the plate as close as I could get it to the first pair of output devices. It’s connected in series with the 9Vac line from the auxiliary transformer to the soft starter and loss of AC detection terminal of the speaker protection module. If the plate gets to 60° (it never should), then the soft starter relays release mains active from the big toroidal transformer, and the speakers are shorted and released from the amplifier instantaneously.
Each of the P68s had already been tested at ±12V (above), but ±70V could blow them up! Well on the first build that happened to one channel through a silly fault of my own. I made triple sure that same error wasn’t repeated here though. So the next test was to make sure that the power supply worked, that the fan comes on, and that the three relays clicked in the correct order. That with rail fuses removed from the P68 main board:
Lights! So far so good. Fan spinning with good airflow, ±66V at the fuse clips referenced to the star grounding bolt (220V mains).
9AWG speaker wires. Red bifurcation for output from the relay, and white for return to the PSU’s top-side output end GND lug.
Slight chatter from the speaker protection relay that I fixed later.
Next step was to add 315mA “test” fuses and hope for the best… All good. Both amps tested fine, so 12.5A fast blow rail fuses were installed and each amp was tested on the reduced mains of 220V to produce a clean 200W RMS into an 8Ω dummy load before clipping a 100Hz sine. So the amps will easily produce 238W with 240V mains and almost double that into the 4Ω load of the speakers thanks to the additional output devices of the extra “half board”. On my home’s 250V mains the amp would do 260/520W RMS with peaks of around 550W into the speakers, but I’ll run them at 230V. I haven’t run REW sweeps through these amps, but they would be the same as those of the earlier P68 build. If interested the graphs are at the bottom of this page.
Aside and another mini-rant about the crap that these amps replace
The bass sections of the speakers were very carefully designed back in 2011 to align Xmax of the Morel drivers with a 200W RMS signal at 20Hz. This was done on the basis of Hypex’s published specifications which I only found much later to be grossly exaggerated. The transformer had 42V secondary windings. You do the maths. The speakers will now perform as intended.
Installing into the speakers
The old amps were IMO fundamentally inept with reversed indications of phase at their phase switches being more evidence of it! 0° was in fact 180°, so I had reversed the speaker wires. This needed to be corrected for the new amps, so while fixing that, I passed the wires through the holes in the correct sequence to line up with a new terminal block:
… which of course meant pulling all the drivers!
That wool damping was purchase from a crafts dealer and was intended for stuffing Teddy bears! 😀
The plates each got an adhesive neoprene tape gasket around the perimeter to ensure a proper seal. These are pierced by the mounting screws:
Speaker tipped over:
In with about 1mm jiggle room to spare:
Well no hum or buzz, some pretty solid bass. Very tight!
Problem integrating with other subs in the room!
Further evidence of what I perceive more and more as gross incompetence from Hypex has revealed itself. Now after replacing the plate amps, the coffee table subs cancelled out the bass! Turning the coffee table sub controller’s level up on any test tone below 80Hz from zero to 12 O-clock took 20dB off the bass (SPL meter C-weighing). As noted above, the old Hypex amps had a phase shifter dial and an inversion switch, but no matter what I did with them, the LR4 integration of the woofers with the mids was skewed! The stupid controls only moved an unwanted “phase bastardisation” up or down in frequency to wreak havoc. And no bypass switch! Nonetheless the main speakers’ bass output integrated well with the coffee table subs by sheer fluke.
With the new amps, the woofers and mids are connected correctly, so hopefully a simple speaker wire flip to the coffee table subs would correct the problem and thankfully it did – well for the most part.
Here is a new unwrapped acoustic phase overlay of the coffee table subs (red) and the front speakers (blue) measured at the seat with a loop-back timing reference for assured accuracy. The coffee table subs have a 36dB/octave LP cornering at 90Hz:
Note: I don’t trust REW’s low frequency phase data. I think the developer even says that it is limited by the broad nature of the impulse peak.
Nonetheless, if it is correct, the phase alignment is good below 80Hz, but there is spread from 80Hz up. So a planned modification of the coffee table subwoofer controller’s filtering will see its low-pass corner shifted from 90Hz down to 70Hz (still good for sub-synth). The coffee table subs are situated much closer to the sofa than the main speakers and the effect of the filters is to lag phase, so lowering the corner frequency will increase lag and so reduce the amount by which the subs lead the main speakers from 80Hz up.
All sorted and sounding great
The plate amps have now been run hard for over 20 hours and after a few tweaks to the up-front bass EQ to in-room measurements, are delivering incredible power to the room. The higher voltage swing gets the Morel woofers moving like never before with very tight control. The speakers now work as designed with in-room response (at the sofa) within ±2dB right down to 10Hz. The up-front parametric EQ needed a 4dB cut at 12Hz! It’s the way things should have been at the outset. The original chicken piss amps never came anywhere close to delivering this level of performance. I suspect that their so-called 12Hz HP filters actually cornered much higher than that.
Yeah. I’ll explain that later.
- Don’t buy subwoofer plate amplifiers with stupid “features” like phase shifters that cannot be bypassed!
- Manufactures’ stated specifications are not to be trusted.
- Commercial plate amplifiers have built-in obsolescence and will fail to an irreparable state much sooner than hoped.