Construction

This project gave the opportunity to practice some more metalwork, something I enjoyed when building the LM4780 Micro-Amp. For some time now, I've been evolving the design of a simple chassis suitable for the high powered discrete power amplifier that I'm planning, and this project would be a practice run for this idea.

A pair of heatsinks form the sides of the amplifier, so the overall dimensions are set by these. Internally, 10/16mm barstock and a 3mm thick piece of aluminium form the chassis. The front, top and bottom panels bolt to the internal framework, and unlike most designs I've seen, do not form structural parts of the case. Extra space is gained beneath the internal chassis using the "false floor" technique.

A key feature of this design is the multi-part rear panel, something I've never seen elsewhere. The ends of the rear panel are formed by right-angle stock which are bolted to the heat sinks. To access the electronics mounted on the heat sink, the heat sink is removed, complete with the right-angle section that carries the rear panel connections and controls. This greatly simplifies the internal wiring, and makes development and future servicing much easier.

Heatsinks

These were recycled from some old kit. The CD here gives an idea of scale. They have a number of holes already drilled in them, and while I was able to make use of the four in the corners, the remaining four remain unused. This is often the case, but given the cost of new heatsinks, it's a fair compromise...

Heatsinks (55K)

Initial framework

The first step was to join the front rails to the angle sections that form the front corners. I wanted to rebate the ends of the bars so started work with a file. I soon realised that this was good exercise, but ultimately pointless so I decided to risk something else - using the router.

I'd read that you can use a woodwork router with aluminium, but had never been brave enough to try it. So after carefully clamping up the work and double-checking everything, I selected my smallest TCT routing bit and wound down the speed of the router to minimum to ensure the lowest cutting speed. With a drop of oil on the bit and decent safety glasses, here are the results:

Aluminium bars (39K)

I worked slowly, taking off no less than around half a millimetre at a time (both depth and width of cut). It was time consuming, but much quicker than filing by hand. I probably could have worked faster, but I was being ultra-cautious at this stage...

Here are the completed front and rear frames. The rear angle sections don't need recessing as the centre section of the rear panel is made from a section of 3mm aluminium sheet... To secure the front angle sections to the frame, two M3 bolts at each corner provide the required rigidity.

Completed frames (38K)

The heatsinks were bolted to the angle-sections using M4 allen screws. Taking great care to ensure everything is true, the dimensions for the chassis could be accurately established. Incidentally, the bottom rails are taller (16mm) to provide space for extra electronics.

Chassis sheetwork

Impressed with the results from the initial routing, I decided to try the technique for the chassis. After roughly cutting the panel to size from 3mm sheet using a jigsaw, I clamped a straight edge to to the sheet, and used the router to cut the sheet exactly to size. The results were superb...

Basic chassis (29K)

You can see that the the front of the heatsinks have been pushed back from the front of the chassis. This increased the surface area of metal exposed to the ambient air, and also means that the front panel forms another heat sink fin - I thought this was a neat detail that I've never seen elsewhere. As well as the improved heat sink performance, it also results in some extra space in the case, which was greatly appreciated later!

Chassis with top and bottom covers in place (24K)

Next came the top and bottom covers. Originally I had intended to use some thin painted aluminium that I had - apart from needing a respray, these would have done the job well enough. But while searching through my stocks of "recycled" junk, I came across the two panels you see here. These have a rather nice brushed finish, along with ventilation holes, and being 1.6mm stock, they are rather more substantial than the original panels.

Believe it or not, these were exactly the right width - I simply had to cut around 2 inches from the front to make them fit. This was worrying - I never have that sort of luck! Of course, if I'd found them first and decided to make the frame and chassis fit around them, there would have been no chance of getting such a good result!

Cutting them down was easy - after roughly cutting them to size with the jigsaw, I attached them to the frame. Then a file was used to remove the high spots which got the panels correct to around a millimetre or so. Next, a sheet of coarse wet and dry abrasive was taped to the bench and the whole assembly was passed back and forth over the sheet until the panels were completely flush with the frame. The operation was repeated with a less coarse grade to improve the finish.

Next, attention was turned to the front and rear panels. These were cut in the same way as the chassis, using a jigsaw for the rough cut and a router guided by a straight edge to perfect the edges. This is time-consuming, particularly the marking out and setup stages, but the results are really superb. You would need to invest in a milling machine to improve on the results. For future jobs, I'm sure I can fashion some jigs to speed up the work, but for now, this was all part of the fun!

Front panel (17K)

This front panel is a 3U blanking panel, cut down on three sides. It is power-coated in a standard off-white colour, which should form the basis of a good paint job. It's 4mm thick (I don't "get" the half-inch thick front panel thing), and secured to the frame with short M5 allen bolts. As mentioned before, it doesn't form part of the basic structure, so it's easily removed for access to the insides if required. As an aside, I've never liked cases where the front panel is bolted to the fins of the side heat sinks - it just looks wrong to me.

Rear panel (20K)

The rear panel is made from the same 3mm stock as the internal chassis, and is exactly the same width.

Heatsink removed (15K)Now that the rear panel is in place, the angle sections can be removed along with their respective heat sinks, as shown here. This was a key feature of my original design, which solves the problems of access for development and service - I've seen plenty of amplifiers that are very bad in this respect. Also, the rear angle-section can hold all the connectors and controls that relate directly to the amplifier, meaning cable-runs can be short and free from unnecessary connectors.

All of the rear panels are secured to the frame with M3.5 bolts - probably overkill, but it doesn't hurt :-)

Heatsink assemblies

I next turned my attention to the LM4780s. Using the same (rather time-consuming) techniques as the Micro-Amp, I prepared the ICs using point to point wiring. But I wanted the best of all worlds here, so the chips are combined with a simple homemade "PCB". This was made using the scalpel and heat method:

LM4780 wiring (27K)

The "PCB" is very simple, and provides a central power supply to the ICs, and holds a few components securely. The ICs are held on the heat sink by a piece of angle, which provides a more consistent pressure between the ICs and the heatsink - the PCB is attached to this bracket, also a piece of Veroboard containing the input buffer is held by the bracket.

Completed heat sink assembly (46K)

This picture shows a completed assembly. As you can see, the LM4780s are buried under the bracket. There is a layer of thermally conductive "Sil-Pad" between the chips and heatsink, and also between the bracket and the ICs. You can see the "hybrid" construction uses point-to-point, PCB and Veroboard, taking the best points of each, in my opinion. Using Veroboard for the front end of the circuitry means that future tweaks and modifications are relatively easy; meanwhile the high power circuitry is very solid and utilises very short wire lengths. Note the connector on the right which allows the amplifier module to be quickly removed when required...

Also note the 10mm bar section added to the top of the heat sink. This was an after-thought to improve the thermal performance - I decided that the top panel should play its part in cooling the amplifier, so the bar conducts heat to this panel. It also helps to increase the strength and rigidity of the enclosure. While on this subject, I've subsequently added a piece of Sil-Pad to the heatsink, such that when the heat sink is bolted to the chassis, heat is conducted to the front angle-section and therefore the front panel. Surface area is essential to reduce operating temperatures.

Chassis assembly

Having built a working heatsink assembly and tested it with a bench power supply, the next stage was to return to the chassis and think about the major components that needed installing. The first item to consider was the power switch. I wanted it positioned in the centre of the front panel, but this meant the switch would potentially clash with the mains transformer. If I hadn't added an extra 10mm to the chassis depth by spacing the front panel from the front of the heatsinks (as described above), the switch wouldn't have fitted.

I also wanted to mount it as low as possible. So I removed a section of the chassis:

Switch opening and insulator (22K)

You can also see the insulator that guarantees that the switch connections can't come into contact with anything else. This is perhaps unnecessary because they are already separately insulated, but as they are at half-mains potential because of the dropper capacitor PSU, I thought it was worth taking all possible precautions. Although the insulator simply looks like a piece of card, rest assured it's a properly rated material - I recycled it from an old switched-mode power supply, and it's a very tough, fibrous material...

Rear of switch (26K)

Bearing in mind the live connections to the switch, I thought that it was important to ensure that the metal body of the switch was connected to earth. Here you can see a small piece of thin FR4 copper-clad material with an earth wire attached. The other end of this wire bolts securely to main chassis. Note also the short length of red heat-shrink which provides additional security...

Careful floor-management was required to fit everything to both sides of this chassis - above you can see the pencil outline of where the mains transformer will go, and around the edges there are various threaded holes for the smoothing capacitors and diode bridges. This picture shows the completed underside of the chassis:

View of the underneath of the main chassis (70K)

In the centre you can see the M8 bolt that holds the transformer. Between this bolt and the front of the chassis is a short length of 16 by 10 mm barstock, which is there to prevent the bottom panel being bent and touching or crushing any of the electronics under the false floor. This also holds the switch insulator in place, as you might just be able to see.

At the rear you can see the bypass relay, with the soft-start resistor to the left of this. The safety earth wire from the front panel switch is visible bottom-right, and the transformer's electrostatic screen shares a screw with the soft-start resistor - this is only acceptable because it's not a safety earth. The two pieces of angle-stock visible either side of the chassis at the front hold the PSU PCB, and the connectors to the heatsink assemblies (click to see these). This all worked out really well in the end, but needed a lot of planning to get right!

For safety, there is an insulating layer that covers all of the electronics:

Safety insulator (48K)

It's securely screwed on - note how the boards are secured with short threaded spacers rather than nuts - meaning that the insulator can simply be screwed to them. Also they are also much easier to work with, as they can easily be done up with fingertips rather than a nutspinner.

As mentioned above, the white insulator between the boards and the chassis is not paper! The separation between the two is only a couple of millimetres - the thickness of an M3 nut - so a good quality insulator is essential.