In 1999 a manufacturer called 47 labs released a 3000 USD amplifier called a gaincard. Although it got good reviews it caused a great deal of controversy because it was based around a pair of 5 USD LM series power amplifier IC's with a basic power supply. So a 3000 USD product had about 100 USD of parts inside. It is not unusual to find any audio equipment selling for at least 10x the parts cost today. Coming from an electronics background I regularly see equipment I would not pay the retail price for based on the grounds the build cost being so low in comparison. In that instance it it best to go the 2nd hand eBay route or build something similar.
Generally most audio enthusiasts disregard chip based amps favouring discrete builds for their superior sound quality. However power IC's like the LM3886 have made improvements and audio DIY builders started to build clones of the original gaincard which resulted in the term gainclone. The gainclone is one of the easiest amplifiers you can build from scratch since the main circuit has very few parts. This is one of its advantages, minimalism with short signal paths on the main circuit and a very low component count. This is a perfect starter DIY audio amplifier build and the sound quality of the finished unit can be comparable to amps up to £1000 with a build cost of about £100.
I decided to build a gainclone since I wanted a compact shoebox style amplifier for the TV system and the fact it looked easy with good feedback from others that had built it.
PCB vs Hard Wired
If you do some searching such as on diyaudio.com you will see that it is possible to build the main circuit hard wired point to point directly on the chip itself mounted on a heatsink due to the low component count. However since there are small ready made PCB's available cheaply on ebay this is a simpler cleaner option. A common gainclone pcb is the XY pcb. This can be bought with a power supply pcb and even with all the main circuit parts needed. I choose this route but replaced a few of the parts with higher quality ones I already had. The XY pcb has a few stability options missing but these are easily fixed and not always strictly needed.
Case and Power Supply
The mains transformer and case are the most expensive parts in this build and in most amplifier projects to be honest. I bought the case from ebay and most of the holes were already punched since the case was designed for housing a small power amplifier. I only needed to drill 6 new holes and enlarge 2 for the RCA sockets so not much metal work which is always the worst task. The mains transformer I used was a RS Components 225VA 230V to 18-0-18-0 and this gives +-27V DC on UK mains which is an ideal voltage for the LM3886 since +-28V DC is usually aimed for. I would recommend 160VA to 250VA in size to power both boards. (AC Voltage x 1.414 = DC Voltage).
Standard PSU caps are 4 x 10,000uF giving 20,000uF per rail. I had a pair of better quality 22,000uF so I used these as the first caps on the board and the 10,000uF as the second caps giving 32,000uF per rail in total. Each was bypassed with a 0.1uF Wima cap soldered on the underside. The original gaincard used a very small amount of supply smoothing and the amount of smoothing can change the sonic character of the amp. Some builders have experimented with this but I just used what I had to get the very least supply ripple and noise as possible. I measured the supply ripple at 27mV on each rail with this setup.
I used a 225 VA RS own brand basic model toroidal transformer which was much cheaper than other brands but it turned out to be noisy mechanically and the single neoprene mounting pad did not provide enough isolation from the chassis base to reduce mechanical buzzing. In the end I used some self adhesive foam pads stuck on the neoprene disc pad placed at the bottom to de-couple the transformer from the chassis which worked quite well. You will get better results with a well known quality brand of transformer. If you do get mechanical noise just mount the transformer on some thick foam or rubber pads on the bottom of the chassis to isolate it. I have done some reading on this issue and it can either be caused by DC on the mains or transformer quality. Since my other amp which uses 2 x 300VA from another brand do not have this issue I put it down to the quality.
Startup Inrush Control
If you plan on using a transformer bigger than 160VA I would recommend fitting an inrush suppressor. I used a 10R 6A type and this solders in series with the primary live (brown) of the transformer fixed to either the mains switch or IEC socket. Its function is to present a 10R primary supply load at switch on. At switch on, large toroidal transformers present a dead short and that coupled with large power supply capacitors initially charging up can result in a lot of current being pulled for a faction of a second which shortens the life of the mains switch. After a while the resistance of the suppressor drops ideally to zero so is almost out of circuit in normal use.
An alternative method is to buy a startup current limiter kit with PCB off ebay if you have the space to fit it. These use a large wire wound resistor and by-pass relay on a timer circuit.
The small XY PCB's are missing a few optional parts for stability. Also some of the DIY'ers over at diyaudio.com have made a track cut to the XY board ground plane to isolate the audio input ground from the common ground plane on the board. You can find all the threads over on that forum. You may not need to do this track cut and in the end this would have been detrimental in my build for the lowest level of hum at 4.5mV per channel. I did not cut the track but used a star earth system as shown below for the main supply grounds and speaker returns.
The pcb's are very small and light so I did not bother fixing the pcb to the heatsink relying on the LM3886 mount. The board is solid and does not move but obviously you cannot put high pressure on it since you would then bend the IC legs. The reason for this is the IC legs are pre bent and designed for it to be mounted vertical on the PCB. This then puts the PCB at right angles to the heatsink and there was no space at the bottom of the sink to secure the boards to the base of the case. I have not found any suitable 90 degree pcb mounts which is what is really needed without fabricating something.
The heatsinks shown here are overkill and you could use a much smaller heatsink a quarter the size inside any case with vents on top. You could then mount the pcb's securely in almost any orientation fixing both the board and the heatsink since it does not form part of the case structure like in my build.
Optional Stability Parts
The pcb is missing a zobel network and a thiele network. These are optional and some people fit them both for good measure. I omitted the zobel network but fitted the thiele network coil and resistor at the speaker output terminals. Some have soldered this on the pcb itself but it works fine at the output terminals and is easier to mount. The coil is made from thick enamelled copper wire 20 turns wound round a pencil or pen. This measures 0.7uH which is the desired value. Using a bigger diameter winding such as wound round a large board marker pen would require less turns, typically 10. The final 0.7uH is not set in stone so some variation is ok. This coil needs to be soldered in parallel with a 10R wire wound resistor. I used a 6W vitreous enamelled type but any wire wound about 3W should be fine. This network provides stability depending on the speaker load and longer speaker leads with higher capacitance.
It is also possible to solder a small pF silver mica capacitor at the input of the amp. This is sometimes fixed between the inverting and non inverting inputs on the pcb but you can solder it directly at the underside of the input terminal block between the input and ground. I used a 120pF, some suggest a 220pF. This simply shunts any very high frequency RFI noise to ground so it does not get amplified. I later removed this since my unit is very stable and I found the small cap was affecting the very highest frequencies losing some sparkle. If you want a smoother top end then fit this cap.
Any HF pops and clicks can be fixed with this small cap but I found that the earthing is far more important and has an effect on such noise esp at switch-on and switch-off. If the earthing is good then the LM3886 is silent in operation with no noise issues, pops and clicks and this cap is not needed.
Fixing Earth and Hum Issues
Earth routing is a subject in itself and can be quite frustrating when you have a finished amplifier with hum issues. You can have a perfect noise free setup on a test board on the bench outside a case, but as soon as it goes in a case everything can change. I would generally recommend using a star earth system as shown in the last picture. This is a metal bolt or screw on the chassis that is also the safety ground. Sometimes the star earth can be isolated from the chassis and route back to the iec earth via a seperate wire from the safety earth. I have never found any hard and fast rules regarding earthing and it is usually a little trial and error at the testing phase. I have found in many builds using what seems to be the shortest and most optimum routing does not always give the best results.
The main supply ground wires from the LM3886 PCB needs to be routed well away from the mains transformer to avoid induced hum even if this makes the wire longer than the shortest path. The speaker grounds should also route back to the star earth using thick wire.
The audio signal wires should be screened unless the case is large so they can be routed well away from the power supply and maintained very short. I have used silver unscreened wire in past builds in tube amps that are inherently more noisy with amazing results in terms of mid transparency, extended clarity and timing. All the screens of the audio cable should be common at one point and this is usually on the volume pot ground pins. Here is when things can get tricky. From the audio common ground point on the pot it should make sense to now route a single earth wire back to the star earth so it has its own clean return path rather than via the pcb grounds. On testing I was getting between 10 and 15mv of hum with this method.
Regarding earthing a rule of thought is to think of earth currents as flowing water returning back to a single point. The strong fast flowing currents are from the power supply sections and speaker returns. The smaller delicate flowing currents are from the audio return. What you want is the smaller delicate currents joining the return path upstream. With this in mind I removed the single return earth from the star coming from the pot earth common. Instead I connected the screen earth of the audio cable to the audio input ground on the xy pcb since technically this is upstream from the psu pcb ground. The drawing below should help clarify this. This dropped the hum quite dramatically to less than 5mV each side, my scope reads 2mv residual noise just from the leads alone.
Therefore isolating the audio ground on the pcb by cutting tracks like some have suggested would have been a backward step in my build. Any value of 5mV or less is only audible with your ear right up against the driver and a good result. If you are getting 2mV with a scope then you are essentially zero since it is almost impossible to read flat zero. If using a multimeter on AC mV scale to take a reading then be aware you will get a wrong reading unless you are using a true RMS meter since the noise at this level is not a pure sine wave.
When fitting the earth wiring I would keep it symmetrical such that the return supply earth from the left pcb is the same length as the one from the right pcb. This is easier if the boards are on opposite sides of the case and the supply in the middle.
I expected the sound to be bright for some reason and lacking bass since I used the quoted 1uF input capacitor. Some builders over on diyaudio.com have replaced this with a 4.7uF but since the input impedance of any op-amp is very high the 1uF is technically enough if you do the maths. A value too high and you are just amplifying sub bass well outside normal hearing. Increasing this to 2.2uF would be adequate. The sound is a little forward before run-in but overall smooth and controlled esp when the optional 120pF cap is fitted. It is not as refined and organic as some class A and tube based amplifiers I have heard. Bass is punchy without sounding bloated which can happen if the input cap is too large a value.
I have experimented with silver interconnects and triple screened OFC cable all with very different measured capacitance values. The OFC cable measured a high 250pF and my DIY silver litz style cable measured 20pF. The latter cable on this amp created a vastly more open and extended sound. Other cables created a richer smoother sound with deeper bass. Therefore experiment with both speaker and interconnect cables since this amp is quite dependent. The interconnect cables with the highest capacitance give a smoother sound just as fitting the optional 120pF cap at the input does. Adding this cap is electrically the same as having high C cables which change the sound.
Supplement RS Parts List
RS 6719038 : 235VA 230V to 18-0-18-0
RS 6719202 : Transformer Mounting Discs
RS 6719218 : Transformer Neoprene Pads
RS 5167849: 10R 6A Inrush Current Suppressor
RS 7760932 : Optional Warning Label
RS 3693447: Chrome 6mm Green LED 24v DC
Maplin Shark Screened Audio Wire
** All Other Parts from eBay **