Ranting in the dark

Actually, this is my first finished project. It's been working for a long time, but since things like those are never really "completed", it has waited a while for its presentation. Here is my wine-box-gone-headphone-amplifier:



Why a headphone amplifier? Well there are many reasons - boredom, curiosity, the fact that a lot of the sound cards on the market just suck horribly bad. Not only did I have a problem finding one that doesn't dump a horrible amount of noise on top of my music but even when I did that, it was too weak to drive my average quality cans. When you want to listen to cruel drum&bass on proper headphones, you need a little more of a punch.

So I figured, it'll be small and cheap to take care of that little problem and on I went gathering parts for what I found to be the most interesting design around. It's all IC-based. Opamps drive high-speed, high-current (for this type of circuit anyway) buffers, which deliver the power into the headphones. The circuit features an actively driven ground "channel" - that is, the headphones aren't simply grounded at the power supply ground, but at the output of an identically designed channel. Since the ground is virtual, a simple rail splitter that divides into power supply rails in two, this is not only better for sound reproduction, but actually necessary. The circuit and board are designed by the awesomely helpful Tangent and can be found at his site here - http://tangentsoft.net/audio/pimeta/ . He actually has an upgraded version of the PIMETA amp, which I intend to try when I get bored again.

The power supply is the funniest part of this build. I happened to disassemble an old modem to salvage useful parts of the circuit. I literally cut out the unit out of the board, using a Dremel. It comes with power line filtering, fusing, nice laminated core transformer, the whole lot. This AC is then rectified, filtered and regulated by a miniature board I also got from Tangent, which provides an amazingly stable supply for its cost and complexity.

The rest of the parts I calculated and sourced myself, assembled the thing and oddly enough it played music from the very start. It was sitting in raw and naked form on my desk while I was working on it and it that form it played music for months. I finally decided that I should hide the dangerous part, the actual AC IEC connector, inside a box somewhere. That's when I found the use for a wine box I had in the office. I checked and it turned out I can fit the boards perfectly inside. That box actually has quite a history behind it, it's funny to think about when I got it and how it ended up spending time on the shelves in my office. Now it features as a headphone amplifier. I hope Mr. John Blandy doesn't mind the sudden product range expansion of his brand.

The casework is notoriously bad. At least when you know how I've done all the details. I had no tools at the time I did it, so half of it is improvised. The end result is surprisingly good, all things considered. It's shiny, it looks very DIY, but I didn't set out to win a design competition anyway.

For the spec curious - the transformer provides me with 18Vac, which I rectify and filter with 6800uF capacitance. This is tightly regulated by an adjustable regulator to provide 22Vdc which is fed to the board. A total of 4000uF of Elna capacitors provide the energy storage for the circuit. Burr-Brown BUF634 buffers can deliver up to 500mA of current per channel, which can basically deliver sound levels that go far beyond destroying your hearing once and forever. Those are operated by Burr-Brown OPA627 low-noise, high-precision operational amplifiers which - I'm being told - sound as fantastic as they come. The ground channel is actually still an OPA124, since I haven't had the chance to replace it yet. The benefits of doing that are very questionable though. And with the price tag of good opamps maybe I'll never do that.

There is one additional advantage hidden in the box, that justifies its constant use. There is an acoustic crossfeed filter at the input of the amplifier. It's role is is to eliminate the exaggerated stereo effect of listening to isolated stereo channels trough headphones. Stereo recordings are awesome and everything but they were made to be reproduced on speakers and your ears and brain have to mix them up together. With headphones the situation is a little different, since one ear gets only one version of the signal. This filter takes care to add the proper mixing and delays in left/right channels so your ears get what they would get in a normal open environment. I have to say - this changed the way I think about headphones forever. I can keep them on the entire day without feeling awkward now. The filter usually has two mixing levels and a bypass switch but I liked it enough to just keep it on all the time.

With my bassy, low-impedance Sennheiser HD212Pro headphones this baby sounds splendid. I can't get the sound card to do such a good job on the sound, period. Not only it delivers much better kick but now I can keep the sound card output in its optimal range, where it drives a high load, at low volume, so there is much less distortion to the signal. Some day I'll just wire this to an external DAC, but that's a story for another post

There. I finally decided that it is finished. I can play with it an tweak it for eternity, but it's done. My 2x50ish watt LM3886-based amplifier.



This baby is a chip-amp, or a "Gainclone", which means it's a single-IC, minimalistic design. It's also a dual-mono design, which means the channels are completely separated, with their own power supplies. it's a power amplifier, so no volume controls or input selections, just pure muscule. The theoretical limit to output power of this IC is 68W, so I've provided more than enough headroom for that. It's powered by two 120VA toroidal core transformers. That's quite a bit more than the 80VA I was planning on, but it did cost the same amount of money, so why not. Now even the craziest bass peak will not trouble my system. With 20 000uF of capacitance per channel, smoothing the output of two rectifier bridges per channel, there's nothing to worry about.

The assembly takes extreme care of making sure that everything is bolted down, glued together and double, even triple insulated. Heatshrink covers every pair of wires. The primary is fused on both live and neutral wires, so are all the four power supply rails inside. To ensure nothing unwanted goes out to the speakers, a DC-protection circuit is installed, that automatically turns off the output in the case of anything wrong with the output signal. It also ensures, along with the LM3886's mute circuit, that there's no output while the amp powers up, as well as the immediate disconnection of the speakers at turn-off.

The chance of ground loops occurring should be taken care of by the ground loop disconnecting circuit, that combines a power resistor and a heavy-duty bridge rectifier to ensure that no unwanted current would flow during operation, but that the immense fault currents will safely flow to ground and break all fuses and circuit breakers, should anything bad happen.

I'm testing it as I am writing this, listening to the amazing Ugress and I can tell you, it was worth any minute I've put into building it. It's small, slick, all-black, with a single white diode mounted at the front of the case. It looks wicked and it delivers a punch to justify its looks.

Simple - possibly quite simple. Awesome - quite a bit indeed. It's so small I can put it in my backpack and carry it anywhere. Yet it sounds fantastic and delivers enough power so that my Cerwin Vega speakers threaten to tear down the house. And despite its miniature size, it doesn't even remotely heat up, even after hours of driving my speakers at painful volumes.

If you've got some spare parts you can build those babies extremely cheap. I did spend some money on transformers and chasis, the major expences related to this project, but you can work around that quite easily without compromising on quality. It still ended up being cheaper than a decent commercial amp, for all it's worth.


Load testing

Well, I've tried and tested the system a lot, during construction and it behaved really well. No one tells you how it would behave after 6 hours of extensive use, wired to two badass speakers and the volume cranked up in party mode. That is to say, stupidly loud, all the time. Apparenly, in small cases, things are expected to heat up.

Heat up, they certainly do. I've had bigger amplifiers than this radiate heat like crazy in such situations. The question was whether the tiny case I'm using can handle all this heat. It heated up, heated up a lot. The cooling flanges on the sides, which are usually only hot in the rear side where the chips are mounted, got heated up all the way. A good thing abount the tiny chasis is that it is well connected to the flanges on the sides, wich means that heat gets transferred across the entire chasis, which seems to help quite a bit. A lot more surface, I guess, even if it doesn't have the best thermal conductivity properties.

There was also one factor that probably influenced temperature in a very bad way - the case has no feet yet, so it was placed on top of a magazine. Otherwise the bottom of the chasis could scratch the surface. The downside of this is that all the air openings on the bottom were shut and there was essentially no convection flow through the amp. I cant bet on it but I'd expect that this airlfow would have reduced the temperature significantly.

The final word on this - it can handle it. No thermal protection kicking in (oh yes, it has that. It's disgusting, but at least it doesn't MELT), no ugly distortions, nor other noticeable effects of a significant increase in temperature. So, hurray, it's passed the final test, what I'd call the "stupidly drunk overload" test. Yey me!