Construction

I used point-to-point wiring with this amplifier - challenging because the pin spacing of the LM4780 is so tight, but much quicker and easier than designing a PCB and getting it etched...

My approach to point-to-point wiring is different to most examples that I've seen, where connections rely on the solder for mechanical strength. This is not conducive to long-term reliability, especially in a power amplifier where high operating temperatures are encountered. To compound the problem, the tight three-dimensional layout that results from this construction technique makes future maintenance difficult.

I firmly believe that all non-PCB joints should rely on a strong mechanical connection and not the solder. So all connections are firmly made by wrapping wires around each other before applying solder. This is much more time-consuming than normal, and is also very fiddly, especially at the beginning. But, as you progress it actually becomes easier than "normal" p2p construction, where the heat from soldering new connections can damage solder joints made previously - in extreme cases, components literally fall off! It's very frustrating when this happens!

As mentioned, I wanted to compliment the design of my Audax mini-monitors by using birch ply for side panels, and also using identical dimensions. I quickly realised that the amplifier could be much smaller than the speakers, so I decided to use the extra depth to hide the connectors. The extra surface area of the aluminium top panel helps with cooling. Ignoring this "overhang", the amplifier only measures 12.5 by 7 by 4.5 centimetres - might this be the smallest Chip-Amp out there?

Assembling the amplifier

The LM4780 has 27 pins, spaced apart by just 1mm. Eight of these are not connected to anything and can be removed. Next, the pins were straightened, and pointed in one of 3 basic directions - horizontally for power, vertically for signal inputs and outputs, and at 45 degrees for ground and mute connections.

For the wiring, I decided to use enamel coated wire. This builds in a safety feature - should the internal wiring be disturbed, the chances of a short-circuit are very much reduced.

The enamel coating will burn away given enough heat from a soldering iron. However, I've not convinced that the resulting fumes are good for your health, and I'm also not happy about the amount of time taken and heat generated by this process. It might be fine when assembling loudspeaker crossovers, but not when performing microsurgery with a tiny chip-amp. So, using a scalpel, I prepared the wires by scraping off the enamel and pre-tinning the exposed copper wire:

Prepared IC and wire (28K)

This next image shows the initial connection. The chip is held in a vice, and the prepared wire has been loosely attached to one of the pins with a tiny dab of solder. Note how all the pins have been colour-coded with permanent markers - just to help prevent any silly mistakes!

Look carefully at the middle two "black" pins, and you'll see that they have been carefully wrapped around the wire, ready for soldering.

First connections (35K)

Here, the two power leads have been successfully attached to the IC. They are separated by a good few millimetres, but the enamel insulation will prevent any nasty surprises.

Both power connections done (29K)

Next I turned my attention to the signal connections, adding the 27K feedback resistors you can see below. These resistors are 1/8 watt - roughly 3mm long. The output and inverting-input pins have been folded back, and the resistors sit on top of the package. Note also how some green sleaving has been applied to the ground pins:

Feedback resistors (32K)

Next additions were the ground wire (same technique as the power connections), and the 1K1 resistors that connect between ground and the inverting inputs. These resistors are bigger 1/4 versions; simply what was available in my parts bin.

Ground and mute connections (27K)

With the ground wire in place, I could think about the components needed on the non-inverting inputs. Before installing these, I connected both the "mute" pins together with some thinner enamal-coated wire, cutting these nice and short so that they wouldn't touch the next resistors. This picture shows the 1/8 watt 1K and 22K resistors. I also added some green sleeving to the non-inverting inputs for peace of mind. Note how the various resistor leads form the input and output connections...

All resistors mounted (28K)

Finally, this view from below shows the power rails and earth more clearly. I was pretty pleased with how this turned out, and at this point was able to quickly test the assembly to prove that I'd not made any silly mistakes. Obviously, it was a bit early for any sound quality judgements!

All done! (29K)

Building the case

This was built entirely from recycled or leftover materials. The sides were formed from the leftovers from hollowing out the layers of the mini-monitors, and the metal was rescued from skips! The smaller pieces of angle-stock are 25mm by 40mm, and the larger piece is around 50mm by 100mm.

Raw materials (26K)

The first step was to make the sides - here are two layers of 23mm birch after being glued together. Much planing and sanding is required at this point...

Side panels underway (56K)

I decided to use my 9.5mm rebating bit in the router to cut the recesses for the chassis and top and bottom panels - this would be easier than making jigs to guide the router. However, this relied on the insides of panels being perfectly straight. This is a good thing, because I like the insides of my projects to be just as perfect as the outside ;-)

Finished side panels (52K)

Here is the chassis, after most of the drilling had been completed. As you can see, the various sockets have been mounted, and the IC has been attached. Note how the sil-pad was left large enough to cover the point-to-point connections - just in case:

Chassis, with IC (41K)

Final wiring of the audio section

At this point, things get very hard to reverse. So I had to be sure the chassis was mechanically finished before proceeding with the final stage of wiring.

This picture shows the star-earth, and the connections to the XLR sockets. Also, we've added the 220uF decoupling capacitors (Panasonic FC's - long-life, high temperature devices). The lead-outs on these could have been shorter but the capacitors would have interfered with the volume potentiometer - if I ever change the pot in the future, I could shorten the lead-outs.

I chose to use XLR sockets for a number of reasons. Most obviously, they are rugged, and of very high quality. But, I was worried about using standard 4mm binding posts when they were so close to the input connectors. This is a non-inverting amplifier, and any stray capacitance between input and output connectors will form a route for positive feedback, with the very real risk of the amplifier turning into an oscillator! The grounded bodies of the XLR sockets offer good screening, reducing the risks. This point is easily overlooked when building very compact amplifiers!

Chassis wiring (52K)

The next stage was to install the input wiring. The cables are high quality video leads, rescued from an old VCR. I made use of the existing termination for neatness. Note the P-clamp that helps to keep the wiring tidy. The black wire is the "pigtail" connected to the screens that connects to the signal earth.

Also, the output wiring as been added, including Zobel components installed on the output sockets. In an ideal world, these would be connected directly to the IC and the star earth, and if this amplifier was any less compact, I would perhaps worry about this.

Input wiring (52K)

Moving forward, the potentiometer components were assembled. The input capacitors and ultrasonic filters can be seen here. The grey 2u2 capacitors are held with double-sided sticky pads, and as you'll see later, their position is for a reason... Note also the signal ground, the bottom-right pin...

Potentiometer and front sub-panel (47K)

Then the sub-panel is secured to the chassis... The signal from the input switch travels via the 1K resistors into the potentiometer. Note the soldered connections on the right side of the capacitors - these are made using the substandard method that I tried to avoid everywhere else - indeed contrast them with the secure connections on the left side of the capacitor. This was for a reason, however. I reckoned that I might need to separate the pot from the chip assembly at some point in the future, so I made these connections easy to desolder. I'm not too worried about reliability here, as it's the easiest to reach when necessary.

Front panel installed (48K)

The final detail was to add the connection between chassis and the star earth, as shown here:

Star earthing (30K)

Installing the thermal sensor/mute/LED board

As the thermal sensor doesn't have to be in close contact with the IC, I decided to minimise internal wiring by placing it directly on the control board. The power LED also connects directly to this board - this is a standard 3mm device that has been filed down to make it fit better into the 3mm front sub-panel.

The 2W power resistor is rather over-specified - the actual power dissipation is around 0.5 watts. But I had one of these spare, and using a large resistor will help reliability by keeping the operating temperature down.

Control circuit (28K)

This small assembly is supported entirely by the toggle switch, as shown here. There is a depression drilled into the chassis for the thermistor to sit in, and a plastic insulating sleeve fitted to the M3 pillar. There is a pair of these pillars which the bottom panel attaches to.

Note also how the grey input capacitors hold the LED in place:

Control circuit installed (41K)

Front Panel

This project gave me the opportunity to try out an idea that I had some years ago - using an overhead transparency film to achieve a professional front panel.

This picture shows the transparency. Looking closely, you might notice a black square around the LED - this is because the opening is 2mm - smaller than the actual LED. One layer has insufficient contrast, so a second layer is stuck behind using a scrap of white tape from my labelling machine which also acts as a defuser to ensure a nice, even light.

Front panel (21K)

And here it is, installed on the chassis along with the perspex which is simply held on by the switches and potentiometer fixings. You can see through the transparency because the toner coverage isn't terribly thick, but this isn't an issue when the case is fully assembled as there is no light source behind the panel.

Front panel assembled (32K)

The visibility of the text is only just acceptable - the perspex is only slightly tinted, but it enough to reduce the contrast. I tried polishing the aluminium to make it as reflective as possible, but ended up applying some white tape to the back of the transparency film. A rather tempting option was to backlight the legends, but this would have required a lot of extra metalwork!