Oct 13 2009
Although presently used to drive a pair of fearsome subs, I originally built this stereo power amplifier for the woofers in some actively crossed 3-way speakers. It’s a wolf in sheep’s clothing and produces an open floodgate of raw low frequency energy without the slightest hint of stress. And that’s into two nominal 4Ω loads. But it’s not for full-range hi-fi!
A big Krell or SimAudio power amp would have been OK, but that’s an unneccessary expense for a mere “subwoofer” application and although capable of driving very low impedance loads, if the actual load doesn’t dip far, the capability is irrelevant. A 400W (say) “professional” or “PA” amplifier might have done the trick, but the cooling fan noise is annoying and apart from Crest Audio, Crown and a few others, I think they’re garbage (especially the Chinese ones with German-sounding names).
The midrange and tweeters needed to be driven by something that sounds decent and which won’t destroy them if a fault develops, so I use valve amps on those.
The bass just needed to be tight and powerful, so I chose standard 400W Class-D modules from the Netherlands. Despite the glowing assessments on some Internet forums about Class-D by random self-proclaimed experts hiding behind pseudonyms and others with clear vested interests, IMO Class-D does for mid and upper frequencies what MP3 does for opera – and it’s not good. Authoritative reports of excellent bass performance however, and very high damping factor figures suggested their suitability for this project considering the bass drivers are wired in parallel.
This photo shows the inside of the finished amp taken before a few modifications were made:
The 625 VA 40-0-40V “Nuvotem Talema” toroidal transformer is a bit bigger than recommended for a pair of the modules, but being a “subwoofer” application, there is no real harm. There is an ESP twin relay soft start for the transformer and power supply capacitors to limit in-rush current when the amp is switched on. It is set with about a ½ second delay.
The filter caps are Panasonic 10,000 µF 100V (40,000 µF total). I might have gone for an SMPS, but hear too many negative anecdotes about them and this was cheaper, might give less cause for unreliability (i.e. toroidal would simply give up rather than explode if over-driven) and was fun to put together – especially cutting that copper ground plate. 😉 More importantly it is heavier and that must be a good thing. 😀 Anyway the amp doesn’t slide away when the power button is pressed! A completed “linear” power supply from the amplifier module manufacturer was an option, but by choosing the transformer, filter caps, bridge rectifier, fuse holders etc. individually you know exactly what you’re getting and have greater flexibility with the component layout within the chassis.
A heat sink is stuck down over those 8 counter-sunk mounting screws with very thing and strong thermal tape. There is heat transfer compound between each module’s blue heat conductor and the aluminium back plate too:
The case has no ventilation slots, so the heat sink is probably a good idea. In use it does get warm so it’s definitely sucking heat out of the modules and the rest of the case stays cool to the touch.
The next photo was taken before finishing all the wiring etc.:
That little bracket in the front is to stop the front panel flexing when the power switch is pressed. The LED is connected across the coil of Relay 1 of the soft start. A 510 Ohm resistor is in series.
A 630V 220nF X2 capacitor across the bridge rectifier inputs to reduce “conducted emissions” passing back down the power cord. The little box (bottom left) is a 10A mains EMI filter.
There is only about 1mm clearance between the capacitor terminals and the chassis lid. There is no potential between the copper plate and the chassis, but the +ve and -ve terminals are at about ± 60V:
So I glued this 5000V dielectric mica sheet to the under-side of the case lid:
There was a small chicken fart from the speakers about 5 seconds after shut-down which was from the on-board OPA2134 input buffers due to each module’s mute function connector being permanently grounded. I added a relay across those afterwards.
Most of the power supply cabling is 10 Amp rated solid copper core with the exception of the earth return from the capacitor ground plate to the transformer centre tap leads and chassis earth which is 20 Amp rated (edit: now removed for ground loop breaker installation – see below). The DC rail fuses are fast blow 4 Amp. The DC cables go through LED holders epoxied into the copper partition for extra insulation. The RCA and Speakon sockets are Neutrik. The bridge rectifier is a 35 Amp 600V Vishay. Like my other recent projects, the chassis is an aluminium Metcase laboratory/medical instrument case from the UK and the ON/OFF switch is a stainless steel Schurter vandal-proof type from Germany. The power switch is one part of the amp that will be touched regularly and if it looked and felt cheap, it would just shout “DIY“! The input signal cable is Teflon-insulated shielded twisted pair with the shield and -ve conductor connected to the respective RCA socket shells. The next picture shows the other end (the green wire is the mute control which goes to ground):
A DPDT mute relay was added with a respective pole across the green wire of each channel. It’s coil is in parallel with that of the second relay of the soft start module, so on powering up the signal is muted until the amp has full power (about ½ a second) and the shut-down noise is muted:
And as a finishing touch, I replaced the ugly case screws with stainless Allan head screws:
At first it seemed to work pretty well for an amateur job. The bass was okay, but as expected mid/upper frequency performance was worse than my 35 year old 50 Watt Teac amplifier!
However, after setting it up in the downstairs system (driving the woofers and with proper input calibration as intended), the bass performance has proven to be quite exceptional.
There was a small buzz in the midrange (through the other amps) which disappeared when the RCA input leads to this amp were disconnected – a ground loop around the interconnects and the power leads of the amplifiers. Some people daringly lift the ground from one or more amplifiers but that is very dangerous. I installed a safety ground loop breaker instead. Here it is alongside the IEC socket:
A 35 Amp rectifier bridge, a 3W 10 Ohm resistor and a 100nF capacitor. It is connected between the right hand end of the copper ground plate at the power supply capacitors and the IEC socket earth pin (and chassis). The power supply transformer’s secondary 0V lines were lifted from the chassis and connected directly to the left end of the copper plate. There is now effectively 10 Ohms resistance between the shells of the RCA input sockets and the chassis instead of a short. The rectifier bridge actually does nothing unless there is a fault in the amp when it would conduct to the IEC ground pin. Very safe. And the buzz has gone. Thanks to the ESP web site for that one. 🙂
I took the amp to an amplifier designer/builder/repairer friend for appraisal and it was found to have an oscillation in the output signal of both channels. It was superimposed over the entire audio spectrum just like an oscillation of a badly executed or faulty amplifier, but here it is an expected remnant of the carrier/switching frequency generated by the on-board oscillators not completely filtered from the output signal by the inductor networks. Although it is of course inaudible and wouldn’t harm even a fragile tweeter, I added a Zobel network across each of the outputs in an attempt to suppress it for peace of mind.
A 7 Watt, 4.7 Ohm (for 4 Ohm load) Welwyn wirewound vitreous enamel resistor and a 400V, 100nF Vishay polypropylene capacitor in series across each output. That ought to quell the mess.
Addendum May 2016
6 years of issue-free use and it has proven to be ideal for its intended purpose. 🙂
Having built and tested a new amplifier, I had the scope and dummy load set up, so I had another look at the output of this amp.
Here is the output with zero input signal:
It’s around 300mV at about 625kHz. This overlays all music output. For instance a low level 80Hz output looks like this:
Of course at higher signal levels it starts to look a lot better.
The hash is of course completely inaudible and although the high frequency energy isn’t much, it must be dissipated by something.