Single Chip 25 Watt/ 8 Ohm Workbench Power Amplifier

Main Index Projects Index

On most workbenches, there is an ever present need for an amplifier.  It usually doesn't need to be overly powerful, but it needs to be predictable, quiet and ready to drive (or listen to) whatever project is currently on the boil.  For many years, my bench amp has been a 3-way active system, with horn loaded midrange (using a home-made Tractrix horn), with a compression driver and horn for the top end.  The bottom end is handled by a dual 300mm woofer in a vented enclosure, which is good for 30Hz.

This is all well and good, but whenever I needed to test a speaker box, I had to drag out one of several power amps and try to find space for it so I could listen to the speaker.  The same applied if I needed to test a speaker driver or some other test (such as verifying the frequency response of a current transformer), whether to verify that it was functional (without making silly noises) or to take measurements.

It was quite obviously well past time to build a dedicated amplifier, with my ever-present BNC input connector, and a pair of combined banana sockets and binding posts for the output.  I use BNC and banana plug to alligator clip leads extensively, and this gave me everything I needed.  It was also obvious that others would find it useful as well.  While it seems so obvious that one should have just such an amplifier on-hand and ever-ready, I had never actually done so.  Yes, I do have a small amp that has its own inbuilt oscillator (in this case, Project 86 - Miniosc), but the amp was under-powered and not as useful as I thought when I built it.  It's also in a case, isn't on my bench normally, and requires an IEC mains lead (plus input and output leads) to be plugged in.  I still had to find space for it on the bench and it was never 'right there' when I needed it.

The simplest way to make a small amplifier is to use an IC power amp, as it's small enough that it doesn't need a substantial case.  If run from a ±23V supply (readily obtained from a 15-0-15V transformer of around 25-50VA), it has enough power for most tests.  Consequently, I decided to use half of a Project 19 LM3886 PCB (stereo is rarely necessary for testing purposes), which has the other benefit that I have a second half-board that I can use for a secondary test amp should the need arise.

Project 19 naturally has the circuit details, but for the sake of convenience it's repeated here.  An input gain control is essential, and the input impedance (even at full volume) should be no less than 10k.  The complete amplifier is shown next, and the power supply is described further down this page.  Input sensitivity is about 600mV for close to full power, so I didn't add a preamp.

Figure 1 - Amplifier Schematic

C2 is shown as 33µF but feel free to increase the value.  This is a bench amp, and good low frequency response can be very handy when running tests.  With 33µF you get a -3dB frequency of 5Hz which should do nicely.  I've also suggested a 50k linear pot, but anything between 20k and 100k will work.  I do recommend a linear pot though, as this is test equipment and it doesn't need to be treated like a hi-fi amplifier with a log volume control.  Indeed, VR1 should be marked as 'Gain' rather than 'Volume'.

The input connector should be the one that you use the most, and have test leads already made up to suit.  All of my bench equipment uses BNC connectors, so that's what I used.  This even lets me use a 1:1 oscilloscope probe.  Note that using it with a 10:1 probe isn't recommended, as they expect a 1MΩ input impedance.  If you prefer to use RCA connectors, that's also fine.  The circuit's ground (chassis) reference should be at the input socket only, with no other connection.

Figure 2 - Amplifier IC Pinouts

The above drawing shows the pinouts for the LM3876 and LM3886.  I recommend that you use the latter, preferably with the 'full-pack' isolated case (LM3886TF), as this means you don't need to use mica insulation or a bush for the mounting screw.  This is a (deliberately) low powered amp, and the full-pack won't impact on performance.  If you happen to have an LM3876 to hand, the PCB accommodates that as well.

Frequency response hasn't been verified above 20kHz (power amps don't like high power at high frequencies), but it's dead flat from below 20Hz to 20kHz.  DC offset measures -1.9mV, so that won't create problems even if I use the amp to drive a transformer.  Output noise measured 75µV (wide band), so it's as close to dead quiet as you're likely to get (-91.5dB ref. 1W/8Ω or 2.83V).  While the power supply shown below includes a mains switch, I didn't include one, so mine is 'always on' like much of my other test gear.  The whole workbench (and all test gear) is turned off when not in use.

The power supply uses a 15-0-15V transformer, and the one I had handy is rated at about 80VA (2.6A at 30V) - total overkill but it was already in my collection.  The capacitor bank I used was also salvaged from another project, and uses 9 × 1,000µF 35V capacitors for each rail.  While a total of 9,000µF per side may seem like overkill (and it is), I figured that I'd rather have a bit too much capacitance than a bit too little.  In general, I'd recommend no less 2 × 2,200µF per side (giving 4,400µF), as the extra capacitance lets the amp run to full power with programme material quite easily without excessive ripple.  Feel free to use more capacitance - that shown below is the minimum recommended.

Note Carefully:  If you are inexperienced with mains wiring, do not attempt to build the power supply.  Have someone experienced wire it for you, and ensure that mains connections are inaccessible to fingers, stray wires or anything else that may place you or anyone else at risk of electric shock or electrocution.  In some jurisdictions it may be unlawful to work on mains wiring unless suitably qualified.  All mains wiring must use mains rated cable and other components.

Be particularly careful to ensure that all mains connections are completely insulated against accidental contact.  Because this is 'workshop' equipment, it may be possible for wires or other objects to come into contact with mains wiring unless it's completely enclosed, either within a sealed chassis or with a suitable cover over the mains terminal block, IEC socket, or however you wire the incoming 120/230V AC.  Workbenches have a habit of having stray wires getting into places where they aren't welcome as other projects are being assembled or tested (especially prototypes!).  You'll notice that an optional snubber network is shown in parallel with the transformer secondary.  For more info on its purpose, see Snubbers For Power Supplies - Are They Necessary And Why Might I Need One?.  The short answers is "you don't" - I didn't include one, and my amp is silent.

Figure 3 - Suggested Power Supply

You will see that I included a 'ground lift' switch, so the amplifier can be fully floating, with no connection at all to mains earth/ ground.  Normally, this isn't something I would ever recommend, but the requirements for test equipment are very different from those for hi-fi systems, and the ability to isolate the amplifier completely is useful.  Don't do anything silly with it though, because the chassis may become 'live' if you do, and that's likely to have an adverse impact on your life expectancy.

You must ensure that the power transformer is up to the task - do not use a cheap unit made 'somewhere in Asia', but make sure it's a reputable brand with little or no chance of electrical breakdown.  The one I used was made by Plitron (not easily obtained here in Australia), but a transformer from any major supplier should be suitable.  If possible, I recommend that you use one with an inbuilt thermal fuse so a meltdown is not possible under any circumstances.  The transformer can be 'conventional' (i.e. using 'E' and 'I' laminations), or toroidal.  A toroidal transformer will be smaller overall, and is a better choice if possible.  In particular, toroidal transformers have a very low radiated magnetic field, and that's important in such a compact amplifier.

The way you build the amp depends on how and where you expect to install it.  Mine is on a 2.5mm thick aluminium 'L' bracket (it's not enclosed) and is mounted below the instrument shelf of my workbench.  The main panel is roughly 190 × 180mm, and holds the power transformer, capacitor bank and bridge rectifier and the power amp board and IC.  The panel is the heatsink, and despite it having been driven hard during some recent tests, it barely gets warm.  I used the 'full pack' version of the LM3886, which doesn't need an insulating washer, but you must use thermal compound (thermal 'grease') to ensure good thermal conduction from the IC to the chassis.

The layout example shown is pretty much how mine is built.  The AC from the transformer connects to the bridge end of the capacitor board (blank PCB copper-clad, 'mechanically etched' with a Dremel or similar).  The DC must be taken from the opposite end from the rectifier, or hum/ buzz will be excessive.  If preferred, simply hard wire the caps together, which is simple and works perfectly.  The DC must still be taken from the caps furthest from the bridge rectifier (physically and electrically).

Figure 4 - Recommended Layout Example

The front panel is attached to the main chassis with a piece of aluminium angle, and that panel has the input, gain control and output connectors (etc.).  The front panel is screwed to the instrument shelf support.  Most DIY people won't have a setup like mine, so you need to work out the details to suit the way you operate.  Since the whole unit only needs to be about 42mm high (internal measurement), it should be easy to accommodate in most setups.  You must make sure that there's enough thermally conductive material (i.e. aluminium) to provide a decent heatsink for the IC.  I have over 340cm², (one side only) which has a thermal resistance of less than 1.5°C/W, and anything smaller could be asking for trouble.  Don't use thinner aluminium unless you are willing to include a commercial heatsink for the IC.  The dimensions shown are those I used, and you don't need to replicate them unless you want to.

One dimension that may puzzle many readers is the distance between the output binding posts - 19mm (19.05mm to be exact, which is 3/4").  This is actually the 'reference' distance, and banana plug to BNC adaptors exist (and are still surprisingly common).  The same spacing is used for multimeter (and many other instruments) inputs.  It's become a de facto standard for test equipment.  You don't have to follow it of course, but I make it a habit as I have several banana plug to BNC adaptors (as shown in the photo) and I figured that they should fit in case I need to use one.  Test equipment always needs to be as flexible as possible.

Most of the wiring is not shown, but one point is important.  The GND output terminal must be wired back to the GND output of the filter capacitor bank, and not to the amplifier.  Also, note that the DC outputs (including GND) must be taken from the output end of the board, and nowhere else.  If connected elsewhere on the board, excessive hum/ buzz may be heard, and this is not desirable in test equipment!

The filter caps don't need to be on a board - for many it will be easier to wire then together and fasten the whole bundle to the chassis with cable ties or other means.  Naturally, it must not be possible for them to come adrift, as a short circuit will damage something (exactly what cannot be predicted).  Due to the (relatively) low average current most of the time, a chassis mounted bridge rectifier isn't necessary, but it does need to be rated for a minimum of 5A (10A is preferred).

This is one of the simpler projects I've published, and (in theory) it doesn't even need an article of its own because everything is described elsewhere.  However, it's also something that doesn't really jump out at anyone as to how useful it can be.  I've messed around with 'full sized' power amps for many, many years, and it only recently dawned on the usefulness of a dedicated bench amp.  I have mine bench mounted, next to my other audio signal sources (FM radio, CD and 'auxiliary'), with a short BNC to BNC lead so I can connect it to my audio output without leads dangling over the bench.  Likewise, my secondary speaker has its inputs brought to the bench as well, so I have everything set up in a way where it's very easy to use.

As mentioned in the introduction, I used the amp to perform a frequency response test on a current transformer (it was much, much better than I ever expected), and the last time I did any similar tests it was a royal pain in the posterior.  While maximum output is only 25W, this is actually more than you normally ever need for a test amplifier, but the extra power may be useful and it lets me do things that would be difficult otherwise.  The low supply voltage (roughly ±24V) on the LM3886 means that its protection circuits won't operate unless I either do something silly (like short the output leads) or try to drive an unrealistically low impedance.

Because the IC has very good frequency response, low distortion and very low DC offset, it's perfect for the job, and I know that if a speaker (for example) sounds distorted, it's almost certainly the speaker at fault, and not the amplifier.  Test equipment is expected to have performance that doesn't compromise test results, and in that respect it's a winner!  Since it uses half of the P19 PCB, the other half can be used to build another, for example if you have two workspaces where the amplifier will be useful.

Project 19 (ESP Project)
LM3886 Datasheet

Main Index Projects Index