My ugly audio amplifier



Here I present my crappy audio amplifier that I built when I was roughly 15 years old. It is half broken now and I plan to take it apart in order to reuse some parts for my current projects. So this article is actually a memorial.

Project presentation

This project started out of curiosity in electronics and audio when I bought a DIY amplifier kit from Conrad Electronic. The schematic of this design is almost like depicted in figure 8-6 of the Hitachi MOSFET application note. Conrad Electronic sold the amplifier kit as HiFi 100W audio amplifier. Of course, it is neither HiFi, nor capable of 100W. The kit included a PCB, all parts for PCB assembly and an L-shaped heat transfer profile. I messed up the PCB and also the profile, but now had invested my pocket money into something that was basically electronics waste. So I had the choice to trash everything or to design and build an new PCB and procure a new profile and start over again. So I designed my own PCB using rub-on symbols on squared paper. This worked reasonably well. Machining the profile on the column drill press of a friend's father, rather than a hand held drilling machine, was successful in the end.

Next obstacle was to find a suitable transformer. I didn't find one with the correct secondary voltage and the one I bought didn't work since the design is absolutely intolerant of supply voltage variations. The voltage was too low. I believe this is when the project went on hold for a while. Later, I bought two toroid transformers that hopefully in combination were closer to the voltage required. Now the voltage was too high. The great benefit of non-potted toroid transformers is that one can easily add windings. So I added some to step down the voltage a bit. The front end of the amplifier now had the correct voltage and the output stage ran on a different voltage. In this case, the OPS voltage was higher than the front end voltage, which doesn't make much sense of course, but at least it worked.

Now came the hardest part: The chassis. Of course, there was nothing suitable or affordable on the market. So I came up with the idea to build everything on a sheet of aluminum and use a ready made 19" front panel. The rest of the chassis was made of wood since wood is cheap and every home improvement store offers to cut it precisely.

The next issue I ran into, was the high inrush current of the toroid transformers. I tried to develop a soft start circuit on my own, but failed. As far as I can remember, the turn-on delay worked, but the instant-off didn't. This was prior to receiving any formal electronics education, so I had no other choice than to buy a time delay relay to control a power relay that bridged the inrush limiting resistor. The power relay also was used to mute the speakers during power up. There was even more control circuity realized in mains voltage logic: A temperature warning and and over-temperature shutdown with higher threshold.

The power supply was completed by a mains voltage filter and C-L-C filters for the DC voltage. Seems like I had a clue about the importance of power supply noise rejection back then. I can remember that I wound the toroid chokes myself. While the basic idea of a C-L-C filter is sound, the implementation probably wasn't. The chokes value is either too high or the filters output capacitance too low. Pretty likely, the filter is resonant.

What I really like is the DIY shielded loudspeaker cable I made. I might even reuse them if possible.

It seems I even had an idea about forming a ground star point, but this idea wasn't followed consequently. It is a miracle that the amp didn't hum like crazy.

Later on, I completed the second channel. The J49 / K134 power transistors became unavailable at Conrad Electronic and I bought another pair nearby in one of the electronic junk shops in central Munich. The price was the same, but they were not genuine for sure. This was my fist experience with counterfeit parts. It was a surprise that the channel built using original transistors eventually died a few years later.

The images show the amplifier in a really bad shape. I encountered issues with crackling noise at low levels and started to remove everything from the signal chain that I suspected of crackling. This is why a cinch cable is hard wired to the amplifier modules, bypassing everything. All non-essential wiring was removed as well. In the end, it turned out that the power relays in the loudspeaker connection was the root cause. Nonetheless, I didn't restore the input wiring.

Schematic discussion

The schematic is the most basic design I can imagine for this dual LTP topology. The design has many serious shortcomings and even contains an error. However, the original schematic from Hitachi is not much better.

The errors first: The Conrad schematic has the value of R15 wrong. This should be in the range of 10Ω, not 10kΩ. In the Hitachi schematic, R16 is missing. While this is somewhat tolerable, the output inductor is undamped and the amplifier will behave in a weird way in case there is a capacitive load at the output. C5 in the Conrad Electronics schematic better should connect to ground instead of the collector of transistor Q3.

Both the original schematic from Hitachi and the one from Conrad Electronics lack a constant current source for the first long tailed pair, which makes the whole design extremely supply voltage dependent and the supply voltage has huge impact on the whole amplifiers bias. This is the reason I had to step down the supply voltage for the front end. A few volts off the nominal voltage the amplifier is designed for, results in a non-functional unit. Both designs also lack emitter degeneration for the long tailed pair transistors, resulting in questionable linearity that is later fixed by negative feedback. This is a poor approach to building a good amplifier.

In simulation with absolutely identical transistor models, this design runs well, but in real life with some variation in transistor parameters, it does not.

There are countless opportunities to improve this basic design using parts worth a few cents. In my opinion, the most dramatic improvement would be a CCS for the first LTP and a bit of emitter degeneration.

For the potentiometer, a multi-ganged type should be used. Conrad Electronic shipped the kit with a single-turn potentiometer, which makes setting correct bias near impossible.

Photo gallery