Building a Power Amplifier

Last updated 23/5/2001

(Note this page probably isn't finished. Hopefully I'll add some more bits and pieces at some point. But there is certainly enough information for those of you who feel confident enough to build their own amps.)

Over the past few years, various people have built various versions of amplifiers which have in some way been based on the Naim 250/135 amps. I myself have sent a load of people a little information 'pack' with a layout I drew ages ago. The amp circuit was based on an old amp schematic which I think originally appeared in 'Audio Conversions' (by Les Wolstenholme I think). I think quite a lot of amps got built from this and they aren't at all bad (although at first there was a slight photocopying error turning a couple of 68 uF caps on the amp board into 6.8 uF which curtailed the bass performance somewhat!).

However it's probably time for an update. Myself and a couple of fellow hi-fi chaps (thanks goes to Jon Nicoll and Pete Wood) have played around with a few bits and pieces and particular components and choices of transistor etc, and the following instructions (in the broadest sense of the word) brings our amp tweaking more or less up to date. We have new layouts, designed by Jon Nicoll (these amps really are 'Nicoll' amps rather than 'McBride' amps), and these are the ones I currently use. And very nice they are too. Having said that, I do not claim that this is an attempt to make power amps 'better' than those commercially available (although in most cases that might indeed be the result); it is more an excuse to get your hands dirty with a very enjoyable, and extremely cost effective way of getting good quality power amps. I will allow myself to claim that, in our opinion, these amps are subjectively very enjoyable to listen to (how's that for British understatement!). Oh, and needless to say, all this stuff is at your own risk.

So cutting to the kill, here are circuits and pcb layouts, and also the what-goes-where instructions.

Circuit diagrams...

Click here for pdf file of the amp circuit diagram.

Click here for pdf file of the regulator circuit diagram.

PCB Artwork...

Amplifier pcb artwork:
Click here for postscript file. Click here for pdf file.

Regulator pcb artwork:
Click here for postscript file. Click here for pdf file.

In case you have trouble printing at the right scale, note that the artwork outer box should be about 76mm x 94mm.

How to populate the boards...

Amplifier what-goes-where layout:
Click here for postscript file. Click here for pdf file.

Regulator what-goes-where layout:
Click here for postscript file. Click here for pdf file.

Components...

If I supplied a full component list, you'd only go through the circuit checking it yourself anyway (or at least that's what I'd do!), and so I leave it to you.

A few things to note though.

On the amp board....

The choice of output transistors (and the drivers) is clearly something you can play with. In the past I have tried various things including BUV20, and then TIP35C devices, but now I would HIGHLY recommend you use the MJL3281A devices from Motorola for the output transistors, and their stablemates, the MJE15032 and MJE15033 devices for the drivers (as shown on the circuit diagram). These are by far and away the best devices I have tried and the amps sounds great and are nice and stable and well behaved with them. They are in the 'flatpack' casing by the way. Use this combination on the regulator board too (ie, MJE15032 and MJE15033 devices for Q106 and Q206 respectively, and MJL3281A devices for Q107 and Q207).

In the past I have successfully used BC550 trannies for Q1 and Q2 although I favour the 2SC2547E devices now (note the pin out on the amp artwork is for the 2SC2547E). I dare say there are many alternatives (noting any change in pin-out of course). Either way, they should be strapped together (back to back with a touch of heatsink compound between them) to thermally couple them. [I hesitate to say it as it may open a can of worms, but if you are able to match transistors, then these two would benefit from being matched].

The little cap (C6) across the ZTX753, I think is actually a a 39 pF in Naim amps, but 47 pF is close enough and is more readily available. Ensure it is a polystyrene cap though. If for any reason you suspect your amps are unstable at high frequencies (high as in kHz or MHz), then a larger value cap here will slug just about everything. Even up to 100 pF can be used (although it does slightly affect the phase characteristics of the amplifier); but you should be fine with 47 pF.

I have just used 2 normal diodes for D1 and D2, although the red LED is a pretty alternative added as an option by Jon Nicoll (who has used it successfully). The LED is arguably better from a technical viewpoint although I wouldn't loose sleep over it.

As the baseline, I like to use the 0.5 W resistors throughout, which is not actually necessary, but gives the pcb that 'chunky feel'!

The resistors R29, R30, R31 and R32 should be 2.5 W versions (eg 2.5 W wirewound). However, I would recommend that for R29 and R30, you use the nice non-inductive type that come in the TO-220 packages (no need to heatsink them). I found that dropping these in did audibly improve the sound.

The resistors R10 (27k) and R8 (1k) are in the feedback loop, and (as with the preamp stuff) benefit from being high quality resistors ie, Vishay bulk foils. I also recommend R3 (4k7) to be a Vishay as this is perhaps most obviously in the input signal path. I found Vishays in these positions do give audible improvements.

An optional little trick you might like, is to replace the R31 (0R22) resistor with a small coil of wire (to give inductance). I have used 1 mm^2 solid core stripped out of twin-and-earth mains cable (keeping the insulation on). About 8 turns, in a coil about 1 cm in diameter is fine. Doing this gives the amps some inductance to chew on, normally supplied by long runs of speaker cable. Frankly I think the amps sound better, and I think it makes them less speaker-cable dependent ie, you can use a lower inductance speaker cable (such as Nordost say).

Regarding the caps... The 10 uF input decoupling capacitor (C1) can actually be a bit bigger. I find a 47 uF audio grade OSCON cap sounds good (Naim use a tantalum) and a N-type Black Gate would be better. The larger caps on the board (C4 and C8) are tantalum types in Naim amps. I have tried Oscons, particularly for C4, although strictly speaking the voltage rating should be at least 35 V (50 V if you are being ultra safe). Having said that, I haven't had any problems (yet!) using a 25 V 68uF Oscon for C4. For normal listening that doesn't approach clipping the amps, it shouldn't be a problem. C3, C6, C7, C10, C11 and C9 are polystyrene types. Note we have added small bypassing caps to C1 and C4. For these we have used polypropylenes. Use a 100 V metalised polyester cap for C14 (220 nF).

A few labelling anomalies on the amp schematics/what-goes-where things...

The schematic refers to an optional R33 (10R) resistor. This is labelled as R34 on the what-goes-where pcb overlay. This was to link the 'noisy ground' NG (ie, that connected to the point marked 0V O/P) with the 'quiet' 0V (ie, that near the input I/P). This was a feature that Jon put in for good reason, but in the end I haven't actually tried it. I connect the '0V O/P' point, and the 0V (near input) point back to the amp 0 V star point (ie, between the psu smoothing caps). The 0 V wire belonging to the signal input from the preamp, runs straight to the 'input' 0 V point on the amp board (ie, not the star point). This all seems to work nicely.

Just to confuse matters, on the what-goes-where diagram, there is a R33, and a C16 which doesn't appear on the amp schematic. This was a leftover from an experiment which didn't work. Leave C16 out (not that we ever told you what it was!) and put a link in for this spurious R33 resistor.

On the regulator board....

Not too much to say, but first off, the circuit diagram still has an older choice of output devices. But it means
Q106=MJE15032
Q206=MJE15033
Q107=MJL3281A
Q207=MJL3281A
Using these means the reg board matches the amp board so to speak, and the pin-outs on the pcb artwork are all consistent.

The caps C101, C201, C105 and C205 are normal electrolytics (do not ignore the voltage ratings!), although I have used decent quality low ESR types (I've no idea whether this is worthwhile but it made me feel good). All the other caps are polystyrene. Having said that, I didn't have 39 pF polystyrenes to hand and so used ceramics for C102 and C202. I would expect 47 pF polystyrenes would be fine too.

Note that Z101 and Z201 are 10 V zener diodes.

The resistors R108, R208, R116 and R216 are 2.5 W wirewound types, and note a labelling anomaly on the reg schematics/what-goes-where things... R116 and R216 on the schematic, are labelled R217 and R218 respectively on the what-goes-where pcb overlay.

The Power Supply...

I haven't yet got round to doing diagrams etc, but really all you need to know is that you want a standard split rail psu. Use a coil which is rated at around 500 VA with 2 secondaries (obviously) rated at 40 V rms. This will thus give you about +56 V and -56 V dc to feed to the regulator board. A couple of smoothing caps in the 15000 uF (and above) range are the things to go for, and I personally like the 22000 uF 63 V BHC Aerovox ALS30 caps (although right now in my amps I actually use 2x10,000 uF caps in parallel; this provides greater ripple current capability, although I don't really know whether it offers any audible difference to the single 22,000 uF approach... single big caps are at least easier to wire up!).

In the psu I use a couple of standard 25 A encapsulated bridge rectifiers. Although I've had success in using discrete Schottky diodes in the preamp supplies, I have never tried them in power amps (the high current ones get pretty expensive!). The bridges can be wired in either 'standard' split rail configuration ie, using all four diodes in each bridge, or just using 2 diodes in each bridge (which is the scheme Naim use).

The true 0 V point for the amplifier is the link between the smoothing caps. It is tempting to sit the 2 caps (assuming you have just a pair of caps) side by side, and run a 'bar' between the terminals, and hang the 0 V wires from this. This is how Naim do it. However Doug Self points out that there are actually large currents running back and forth along this connection, and it is better to connect an additional short bar to make a 'T' shape, and connect the 0 V wires to that (ie, the 'sticky out' bit).

Mains switches should be the best and the most robust you can get as the inrush current at turn on is 'big'. However, there are good 'high inrush' switches readily available rated at 16 A nomimal and 150 A inrush. I use these and have never had a problem with them. I haven't tried any slow-start devices, but they seem like a good idea.

Some general notes...

Building a pair of 135-style monoblocks isn't that difficult, although it's probably only a good idea if you think you'll enjoy doing it (not just because you want cheap amps) and the information on the page makes some sort of sense to you! Needless to say, compare the pcb tracks with the circuit diagram so you know exactly where you are and what links are required etc.

A lot of the expense is in the psu, and some cost is thus saved by building a 250-style amp ie, Naim 135s have a power supply per channel, whereas the 250 powers both channels in parallel from the one power supply.

I attacked the heatsinking by getting a rather hefty 0.4 deg/W comb section (200 x 250 mm) which I cut in half (one half for each monoblock). This allowed me mount both the reg amd amp boards flat on the back of this section (ie, the monoblock heatsink section is 200 mm wide and 125 mm tall with the fins running vertically). In fact, I used the heatsink as the front on my amplifier (the other sides being sheet aluminium) as they are quite pretty anodised black comb, and it allows good airflow. You may choose a different route of course. To be honest, unless I'm playing tracks with big bass on, at very silly volumes, they never really get hot, which suggests heatsink overkill. I did a friend's amp which has everything (including a more modest the heatsink) inside an off-the-shelf 19 inch rack case, and although the airflow over the heatsinks is impaired slightly (although the case does have vents in it) it still never seems to get beyond luke warm.

The output devices are obviously mounted on the heatsinks as a matter of course. It is also advisable to mount on the same heat sink as a pair of output devices, the MJE15032 and MJE15033 devices too, and also the little ZTX384 small signal transistor (the pcb is designed to allow you to do this easily). This little transistor ideally wants to be thermally coupled to the output devices which would make the amp more stable when being driven hard (so I'm told)... although I think you needn't loose sleep over it. If you do not want to mount the MJE15032 and MJE15033 devices on the large heatsink, then at least ensure they have their own TO220 style clip-on sink. The artwork supplied here has all the driver/output devices nicely at one end of the board. I have the pcb mounted on 10 mm spacers, with the output device legs threaded into the pcb from below, allowing plenty of room to twist the devices flat (parallel with the pcb) in order to be pushed against the heatsink face. The devices are secured by a simple metal 'bar' (held by a bolt tapped into the heatsink) which runs along the back of the devices (a little packing with tin foil on the thinner driver devices is needed). To electrically insulate between the devices faces, and the heatsink, I prefer the sheets of thermally conductive 'stuff' (eg, "Kool-Pad" or "Sil-Pad" or "Q-Pad" etc, which can have thermal resistances of down to 0.2 deg/W) rather than washers and heatsink compound (which is messy).

The 1K trimmers near the input of the regulator board set the current trip value. This is a very useful facility as it protects against most mishaps. I believe Naim set their reg boards to trip at around 10 A. Assuming you don't have the facility to do this properly (and I don't) then a rough 'half-way' position for the pots will suffice. If you set them almost fully clockwise, the regulator boards shut down too easily (eg, at turn on!), and fully anti-clockwise means regulator components will blow before the thing shut down. Half way (or at about 1 o'clock perhaps) appears to be about right. If the reg board does shut down at any point then power down the amp and leave it for half an hour or so, and turn it on again (and hopefully it'll work again!).

The other trimmers on the reg board are to set the output of the regulator. The positive voltage rail should be set to +39.2 V and the negative rail to -39.8 V. Note that the voltages drift when the components are new, and also when they are cold just after turn on. So you want to have the amp sitting there for half an hour or so before doing your final setting. And also when you've run the amp in a bit, set the voltages and bias again.

To set the biasing of the amp, do this: initially before you power-up for the first time, set the 2K trimmer on the amp board to about the mid point. When the whole thing is up and running, you want to set the bias by measuring the voltage across R29 (the 0R22 resistor that is attached to the 'positive side' MJL3281A output device emitter). With no input, twiddle the trimmer until you get about at least 2.2 mV (ie, 10 mA bias current). The desired setting is arguable. I find about 20mA (4.4 mV) is about adequate, although I tend to run at 30 mA to be sure. Naim also use about 30 mA with their devices. There's room for experimentation. If you go too high the amps risk thermal runaway, but if heatsinking is adequate then 30 mA is worth using.

And now for a fun bit. Those in the know will notice that our amp circuit diagram does not include the simple protection circuitry which bona fide Naim power amps have. These components would offer some protection in that when the current through the output devices went above 10 A for a significant time, these components would shut the output down. However in my (and everyone I know who has tried it) opinion, this circuitry significantly degrades the sound. Without it, the amps are more powerful sounding, more transparent, and more involving and musical. So we have dispensed with the circuitry. But in our experience, the reg board current trip offers a very good level of protection, so no big deal.

Obviously building amps, with fairly hefty voltages, and large electrolytic caps about, is not trivial and I take no responsibility for any mishaps. On the other hand, it is great fun and the results are really very good.

And that's about it.

However, seeing as this has been a team effort, perhaps some comments from Jon and Pete would be helpful too. So...

Pete Wood's comments on the amps

Jon Nicoll's comments on the amps