|8 Channel Amplifier|
The actual construction of the complete amplifier went pretty straightforward. Probably because I already had built several prototypes in the previous months, and had already built the powersupply (which I used in testing the prototypes). I still took several evenings, simply because of the amount of work involved.
The PCB's were made by Olimex. They're nice people, cheap and accept eagle files directly (no need to export them to gerber files). I checked them thoroughly and discovered a short in one PCB, which was fixed with a knife in no time.
All eight boards were stuffed with all the parts. I have put some very nice (and expensive) caddock non-inductive to220 resistors in these amplifiers, but the normal ceramics probably work just as well (and also fit in this board). They do make for nice pictures though! Most of the parts were bought at Reichelt. One of only a few suppliers that charges affordable prices to private persons. The bias-transistor is thermally coupled with a driver transistor for better bias-tracking and to prevent thermal runaway. (click the image to see what I mean)
The backpanel has standard cinch connectors for inputs, and 10A 4mm connectors for outputs. Eight channels actually is quite a lot of connectors to give a proper place on a panel, even when that panel is large.
When I soldered all the wires to the boards, I got a little nervous. There is no way I can give all these wires a proper place in the chassis. Each board has 10 wires, 8 boards gives 80 wires. Spaghetti is a good description.
And this is the end result. The chassis is from an old 100V PA amplifier, which cost me two bottles of (cheap...) greek wine. The transformer is a second hand amplimo 78012 300VA 2x12V. This was a nice find. It provides enough voltage and adequate current (8 channels need loads of current). It is a simple unregulated powersupply, with around 30mF of capacitance, and a 30 amp bridge rectifier, bolted to the chassis for cooling. I kept the powercables as far away from the signal cables as possible. The heatsinks are simply bolted to the chassis. And I used the original fuse holder and powercable and switch that came with the chassis. The turn on delay is not present in this image, but the amplifier does have one now.
In the partslist on the previous page, R5-R8 are chosen in such a way that the total gain of the amplifier is 1x, which is enough when connected to my soundcard. C20 is 1nF and R34 is 10K. C20 needs to be as small as possible, so that it doesn't limit the slew rate of the amplifier, but can help stabilize the amplifier. There are actual connections for the feedback. Loudspeaker wires have a small resistance. When a large current flows through them, there is a small voltage drop over the speaker wire. When feedback is taken in the usual way (the separate feedback wires not connected), the feedback network "thinks" that the speakers receive a higher voltage than they actually do. By mounting a separate feedback wire (which only has a very small current flowing through it), the feedback network can do it's work better. It is possible to take feedback directly from the speaker terminals on the speaker, or just from the output connectors on the amplifier. Some care must be taken though. Connecting the feedback wires in the wrong polarity, will destroy the amplifier in no time and when the feedback wires run close to high current cables, they can do more harm than good. In my amplifier, these feedback wires are not connected.
The top-layer of the board has a copper-pour, with it's own solder pad (RF), which is connected to the RCA ground terminal. Both inputs on each amplifier board are also connected to the RCA inputs. The outputs to the speakers are also both taken from the amplifier boards. The star ground of the whole amplifier is located on the power supply board. Each board has it's own ground wire to this board, and the chassis is also grounded.
With all these precautions in place, each of the 8 amplifier channels is dead-quiet. I even connected them to some fairly sensitive speakers, and with my ear close to the speaker, I couldn't hear a trace of hum.
The first time I switched the amplifier on, I checked for obvious problems, like overheating components, and oscillations, but luckily everything was fine. Setting the bias and DC-offset were next. DC offset was simple, just turn pot R10 so that DC voltage on the output equals 0V. The output transistors need about 50mA of bias (and when using 0.5ohm emitter resistors is simply 50mV over both resistors). It turned out that I couldn't set the bias low enough. So R13 was changed to 3k3 ohm, and this cured the problem. The amplifier was now left running for some time, to see if anything heated up, or if the bias value changed. None of this happened.
Well, let's hook up a CD-player then! Oh heavenly bliss; finally, it works! From start to finish, the whole project took about 6 months, of which almost half were spent waiting for parts and boards. And now finally, the completed amplifier was playing music. I spent 2 or 3 hours just listening, enjoying the music and getting used to the fact that it was done.
When the listening session was over, I did some measurements on the amplifier, to check how it performed in a somewhat more objective way.
The powersupply supply rails are +/-17V, the amplifier clips as expected at above 30V on my scope screen. And this doesn't change much with a 4 or 8 ohm load.
The picture on the right shows a 20khz square wave, the one with the slight overshoot is the amplifier output. Square waves over 100khz are no problem, with a resistive load connected. Square waves into a capacitor parallel to a resistor show little ringing below 30khz. There was also no crossover distortion visible on the scope screen, when looking at a sine-wave.
I also did noise and distortion tests with sample champion and a delta-410 soundcard. These tests were a little more difficult, because a soundcard doesn't have protected inputs. I kept the voltages below 6V p-p, so that the soundcard didn't get damaged. This doesn't test the amplifier at full power, but it does test the amplifier in the range it gets used, because I will use that same soundcard when designing crossovers. It turns out that the amplifier outperforms my soundcard. When I do a loopback test with my soundcard (connect the output directly to my input), signal to noise ratio varies between 88 to 89 db(A). And THD measurements are around 0.002% (both THD and THD+N). I get the same results when I test the amplifier, which means that I need better equipment to test this amplifier. It also means that although I couldn't measure the amplifier properly, it does perform pretty darn good.
This amplifier is configured for use with my soundcard and for listening to crossovers when designing speakers. The output of my soundcard is high enough to provide an adequate listening level, and this means that each amplifier channel doesn't have to provide a lot of power. Even though the +/-17V supply rails mean that the amplifier can deliver around 15 watts into 8ohm and double that into 4ohm, the 1x gain means each channel would never have to supply more than a few watts. But since this little amplifier board performs so well, I'm planning to use it at more opportunities. One of the more obvious off course are active speakers. The small size is ideal to put multiple channels into one chassis and leave room for crossoverboards. Some tweaks are necessary then, but only in parts values, not in circuit layout. So just as I finished this project, ideas for new ones are already being born. Lovely hobby, this is!