P3A Power Amplifier

During the summer of 2012 I built this amplifier as an upgrade over my existing receiver. It's based on Rod Elliott's P3A design and delivers about 60W into 8 ohms, or a little over 100W into 4 ohms (or it should do, I've never actually measured it). I bought the ESP P3A board, as well as a P33 speaker protection board, and built a P23 clipping detector circuit and a stereo LED power meter onto veroboard.

Parts arriving

Here are some of the parts as they arrived. I'm not using boutique components or any of that crap, but I am using metal film resistors for lowest noise, as Rod Elliott recommends, and Nichicon capacitors, as I want to avoid the cheaper brands mainly because they tend to fail prematurely. I've used Nichicon caps when refurbing other gear and was very pleased with them, but they aren't sold in the UK so I had to order them from Mouser in the US. Fortunately the shipping is free and it turns out Mouser sell most of the resistors and transistors I need for cheaper than I can get them in the UK, so the slightly longer wait for them to get here is worth it.

Resistors, capacitors, transistors, etc.

Some more parts in the box that the chassis came in

Rod Elliott/ESP P3A and P33 circuit boards

The oversized steel chassis

The monstrous heatsink

At first I thought the heasink was oversized—at 0.41K/W it has a much lower thermal resistance than the design calls for, but was actually cheaper than some smaller heatsinks. Nonetheless the thing still gets burning hot when running 4 ohm loads at high power—so much so that I now have a fan sitting behind the amp—so I'm not so sure if it is too big after all. The chassis definitely is too big though, coming in at 4U rackmount size. This was simply the only sensibly priced thing I could find that was big enough. A smaller 3U case would probably work too, but would leave barely any room around that monster of a heatsink.

Completed boards

The first picture below shows the nearly complete P3A and P33 boards. The P3A board (the larger one) is still waiting for a few capacitors and its bias trimmers, which were held up due to a stocking mistake at the supplier. The output transistors are not yet attached, and will wait until I'm ready to put the board on the heatsink. If you look very carefully you might notice that the P33 is also missing one part: a diode in the bottom-left corner. This is because the design allows for two different flyback protection schemes—a single diode on the board, or a network of diodes and zener diodes soldered to the relay pins. I opted for the latter.

Near complete P3A and P33 boards

Clipping indicator and meter driver, with meter schematic

Close-up of clipping indicator/meter board

The second two images (above) show the P23 clipping indicator and LED meter driver circuits built on a piece of Veroboard. Rod Elliott doesn't sell boards for the P23, so I had to design it myself. The clipping detector circuit is up top (mostly discrete components with one 8-pin dual opamp chip) and the meter drivers are at the bottom, surrounding the two larger LM3915 ICs. There are two identical meter driver circuits, one for the left channel and one for the right, and they are almost exactly as shown on the datasheet. The two blue trimmers and supposed to be for calibrating the meters, but they don't quite give the range I expected. As of the time of writing the meters remain only roughly calibrated. The TO-220 device on the left is an LM317 voltage regulator, which reduces the 35V main supply down to 5V to run the meter circuit, getting very hot in the process. This isn't the most efficient way to do the job, but it is by far the simplest. Shame it requires a whole extra heatsink!

Drilling holes in its face

Not quite as savage as it sounds—how else am I supposed to get the LEDs and power switch to fit? :D Getting all 23 LEDs in place wasn't a trivial task, and I was a little concerned about getting all of the holes to line up straight. I decided to model it in Google SketchUp first, then print out a template for drilling. The SketchUp files are available for you to download, but this was the first thing I ever designed in SketchUp (or any CAD software), so the model isn't that good. The large grey circle is the location of the power switch. The blue dot above it is the power LED, with the red fault and clipping LEDs to its right. The two meter towers are towards the center.

Front panel layout

3D model of the amplifier

Plan of front panel

Plan of rear panel

It came out okay. Despite all my planning, what I stupidly didn't anticipate was the drill slipping quite as much as it did. This is where a drill press would be really useful. At a glance, the LED meter towers appear to be dead straight, but you don't have to look too closely to see that they are ever so slightly off. I did think about having a small subpanel machined just for the meters (companies like Schaeffer A.G. will make one-off panels, as will many local machine shops), but I'm happy enough with it for now.

Once the front panel was done, there was still plently of metalwork to be getting on with. I had to drill and cut out holes in the rear, bottom and side panels, including a very big one for the heatsink. The heatsink sits flat on the back of the amp, with a hole in the rear panel exposing most of the heatsink surface to the inside to mount the power transistors on. I also cut out three smaller but still fairly big holes for the power, input and output connectors. None of the audio connectors mount directly on the chassis, instead mounting on small plastic subpanels cut from the case of a broken laptop power supply. This insulates the connections from the grounded chassis, ensuring that all audio circuitry is completely isolated from the mains to prevent hum loops. In the case of the speaker connections, this also stops the positive and negative contacts from shorting. The photos below were taken after the metalwork was complete, the chassis had been resprayed to cover the inevitable damage to the paintwork, all of the connectors, switches, fuseholders and LEDs had been fitted, and some of the electronics had been installed. You can also see the additional internal heatsink for the voltage regulator.

Finished front panel

Some of the electronics have been installed

Biasing and testing

The picture below shows the finished amplifier board mounted on the heatsink. The trimmers and capacitors have all arrived, and I have added the output transistors, which are installed on the underside of the board, flat against the heatsink. You can see a series of holes that have been drilled and tapped around the perimeter of the heatsink, for bolting it to the chassis. There are four more holes you can't see, for mounting the output transistors. They are positioned beneath the screw heads along the top of the board. Tapping the holes was a learning experience, having never done it before. At one point I managed to snap one of the taps by applying too much pressure and had to start a new hole. That's why there are two holes in the top-left corner—one of them has half a tap stuck in it that no amount of drilling could remove! Fortunately the remaining end of the tap still works, as I don't have any others of the same size and thread pitch (and it was from a brand new set of taps and dies I had just bought that day).

Circuit board and heatsink

If you think the board is fitted wonky, it isn't. That's just the perspective of the photo. I may not be the best photographer in the world, but you can bet I make sure all the lines are straight before drilling. The pictures below show a view of the inside of the amplifier, and of a multimeter showing the bias current has been correctly calibrated (actually the meter is measuring the voltage across the emitter resistors). Nothing besides the amplifier board, power supply, and inputs and outputs are connected yet, and the speaker protection relays are still missing.

P3A board is now installed

Setting bias current

First listening test

The last picture shows the first proper listening test. I had briefly rigged the thing up in the garage to make sure it made a noise, but I didn't listen much as the garage isn't the most comfortable place to sit around. So I hauled the whole thing into the living room to give it a proper listening to. None of the LEDs are connected yet so you can't see that it's switched on, but it is. The picture shows a pair of cheap speakers that I used for initial testing—at this point the amp had only been switched on for a few minutes at a time, and I hadn't yet wired up the protection circuits. After I was confident that everything was working (or more like after I got fed up with the crap speakers), I brought my decent speakers out to hear how good it really sounded. There's no preamp in use here; input comes directly from the USB sound card on the right. I mostly played lossless rips and CDs, and the sound was pretty sweet.

Finishing up

After I was satisfied that the amplifier was working, I just had to wire up the LEDs and protection circuits. Here you can see that the protection relays have been glued into place, the wiring has been soldered up and cable ties have been fitted to make everything neat. This is not just for appearances by the way (if it was I wouldn't have bothered). Having the cables neat helps reduce the noise levels. In particular, the AC power cables are kept as far away from any critical wiring as possible, and the input and speaker leads are well seperated. The wires going to the metering LEDs are nowhere near anything sensitive, to keep switching noise out of the audio. Overall the amplifier is almost silent with no signal, there is no noticeable hum, and hiss can only be heard with my ear right next to the tweeter. The majority of the system noise comes from other pieces of gear. And in case you are wondering, those red and yellow wires on the left are not in contact with the mains wiring. There is a good inch or more of seperation vertically, so the whole thing meets safety standards.

Inside the finished amplifier

Bonus points if you can figure our what's playing!

Behind the front panel

The last photo above shows how the LEDs are fitted. All the LEDs are soldered to one of two circuit boards—one for the meters and one for the indicators. The two boards are bolted to an aluminium angle bar which runs across the top of the front panel, secured to the sides with a couple of cut off angle pieces. There's one other thing missing from these pictures: after I had taken the photos and thought I had everything finished up, I discovered that the amplifier had an annoying habbit of blowing its mains fuse every so often at power up. This turned out to be due to the large inrush current drawn by the toroidal transformer. I ended up fitting a soft-start circuit, based on Rod Elliott's P39 but built on veroboard, with a large 50W balast resistor bolted to the chassis. I didn't think to take any pictures of it, so sooner or later I'll have to crack the thing open again and snap a few photos.

Here are a few photos of the finished product. Again, some things look a little wonky in these photos, but that's just due to camera angle (honest!).

Front view of finished amplifier

Read view of finished amplifier

Amplifier installed in my system

Another shot of the complete installation

The last two photos show the amplifier installed in my system. I'm using my old receiver as a preamp, and as a secondary power amplifier to drive another pair of speakers. The receiver didn't come with with a preamp output or power amp input, so I modified it to work with an external splitter unit I built especially (the black box with the two knobs on the front, at the right of the photo). Sometime soon I'll write up the splitter build too. And if you're thinking the whole thing looks odd with the power amp on top—I agree. Unfortunately the receiver is slightly deeper than the power amp, so that's the way it has to go. However, all of this has changed again since the photos were taken. The receiver has now been replaced with a small mixing console and the splitter has been retired, despite only being in use for a few months. The system has also been moved to a different part of the room.