BUILDING THE MJR-7-Mk3. Part 1.
A few of the component values have been changed to ensure they are easily available in the physical sizes needed to fit on the board. With a 60V supply the maximum output into 8R is about 30watts, and 50watts into 4R. If higher power is needed this can be achieved by increasing the supply voltage, then reducing the original onboard 1000uF 63V supply capacitor to 470uF rated at 100V to keep a similar physical size, and increasing the 220uF in the 'anti-thump' circuit also to 100V rating. This increased supply voltage is not recommended for speakers with low impedance, falling much below 5 ohms at any audio frequency. Using the more expensive dual-die mosfets would give a higher maximum rated output current, but part of the problem is heat dissipation, and the dual-die is of limited help with this. The output capacitors only operate at half the supply voltage, so could still be 63V, if we accept the risk of damage under fault conditions. Personally I would increase these also to 100V, but then the physical size becomes a problem. Another pair of holes have been added to the pcb to accommodate 25mm diameter capacitors instead of 22mm. One option is to reduce the parallel pair of capacitors from 2200uF to 1500uF each, the lower values being perfectly adequate if speaker impedance is not too low, a requirement mentioned earlier for higher power use. The output electrolytics I am using are Panasonic TSUP series 2200uF 63V which are 22mm dia and 30mm high, each with current rating 2.52A, giving total 5.04A, which exceeds the 7A peak specification of the mosfets for sinewaves at least. I had guessed that the current rating would be higher for high temperature or low esr electrolytics, but comparing specifications there seems to be no obvious link. The current rating of electrolytics is an average value, and short peaks above this should be no problem.
Before deciding on the higher power option I suggest checking carefully what output level is really needed before worrying too much about power rating. There are many people happily using 5 or 10 watt class-A amplifiers who feel no need for more power. Remember the logarithmic loudness effect, doubling the power is nowhere near a doubling of perceived loudness, and the difference between 30W and 80W is not as dramatic as we might expect. The MJR7 has been designed primarily for quality rather than quantity, and if much higher power is really needed there are other more suitable designs.
The latest circuit is shown next, with the component values I used myself. Only a few component values have been changed, and the original values work just as well if available. The 470p and 330p at the input can be increased to 560p and 390p as originally specified, but don't just change one of them, the ratio should be kept about the same.
Next is a full resolution image of the latest PCB (Feb-20-2009), as viewed from the component side. The blue lines are at 1/10 inch (2.54 mm) intervals, the board size being about 4 x 5 inches. Boards are commonly available in sizes 4 x 6 inches and 100 x 160 mm, and either can be used, just one cut is needed to reduce the length. (The full size can be kept, possibly leaving an unused strip at each end which could be drilled and used for mounting the board in a case instead of my own method of using the heatsink bracket to support the board.)
There are several ways to make the boards, two suitable for DIY constructors are described, the first using a UV light box, the second an etch resist pen.
This layout can be printed on transparent film (I use tracing paper myself, Goldline 112gsm, with a Canon S600 inkjet printer, which gives a good opaque image, but takes a while to dry. I use the high quality, plain paper, greyscale print settings.) The transparency can then be used with an ultraviolet light box and photo-sensitive boards to produce the PCB, a typical exposure time is 3 minutes followed by about 30 sec development time. This is not too difficult, and the equipment need not be expensive. I bought my own light box second-hand on eBay. A small hand drill is adequate, but for more frequent use a good drill with speed control and stand is invaluable. I have the FBS 240/E drill and MB 140/S stand made by Proxxon. A good article about making PCBs can be found here. I used MGI Photosuite to print the image the correct size, this has a 'print preview' which allows adjustment of the printed image size and gives a numerical print size which can be used with a little trial and error to get the right printed dimensions, i.e. a board size 4 x 5 inches. If using this program remember to check the 'aspect ratio' box when adjusting print size, then height and width will stay in the correct ratio. The transparency is used on the UV box with the printed side up so that the printing is in contact with the UV sensitive side of the board to give the clearest image. When developed and etched the name 'MJR7-Mk3' should be printed the right way.
An alternative method avoiding the light box is to print the layout on paper and stick this onto a piece of plain board, e.g. with double-sided sellotape or some other non-permanent method, then drill through the diagram and the board. After cleaning the copper surface use an etch resist pen to draw in the connections. Printing a mirror-image of the board diagram above to copy from helps. Then etch in ferric chloride solution, then clean off the etch-resist, e.g. with 'wire wool' cleaning pads. I sometimes use this method to make a single board, but of course the UV method is neater and quicker if many boards are to be made. An example of this method is shown for an earlier board layout here. I usually drill from the non-copper side using this method, but then there are raised ridges round the holes on the copper side. A better idea may be to use a layer of cardboard, about 1mm thick, on the copper side and drill through this to stop the drill slipping around too much on the copper, but I never tried this yet.
All holes can initially be drilled at 1mm, then the few larger sizes are easier to drill accurately. The holes for the mosfet fixing bolts are 3.2mm or 1/8 inch. The Panasonic TSUP 2200uF output capacitors need 2mm holes. The only other size needed is 1.2mm, which is needed for the presets, the fuse holder, the inductor, and the mosfet leads. The 1mm drill bits should preferably be tungsten carbide if fibreglass boards are used, but cheaper 'HSS' (high speed steel) are adequate for the few larger size holes, although they will soon become blunt. The carbide drills are rather fragile, and liable to break with a hand-drill unless a good stand is used. It takes a little practice to avoid breaking drills, I broke 3 drilling 5 boards, but then did another 10 without breaking any. The type with a thicker top section are apparently easier to break than those with a constant diameter, which can flex a little more before breaking. Using the laminated type boards with just a thin surface layer of fibreglass the drills will last longer, and the HSS type may be good enough if only one or two boards are being made. They are certainly a lot cheaper, though reground carbide drills are available for about £12.50 (UK) for a pack of 10.
Adding the components. Viewed from component side.
The heatsink bracket should be drilled for both mosfets and heatsink before fitting components to the board, then the board can be used as a template to ensure the mosfet mounting holes are in the right place. The bracket is to be positioned as shown next, aligned with one edge of the board, and with the inside back of the bracket aligned with the back of the board, i.e. with a 5mm overlap, and so that the holes for the mosfet leads are about 3mm from the front edge of the bracket.
The following layout diagrams are based on an earlier board design, but there are only small changes to adjust for different component sizes, for example some of the resistors are now specified as 3watts and the types I can get have length 15mm. I have since found some smaller 2W, so this change was not really necessary. High power rating is not essential for these resistors under normal operation, only for high frequency testing or fault conditions. The 2u2 input capacitor has lead pitch 7.5mm, but there are a few similar components available with different lengths.
The 220R preset should initially be set to minimum resistance, i.e. fully anti-clockwise, preferably checked with a meter to be sure the resistance is small, otherwise when first switching on the completed amplifier the quiescent current could start off dangerously high. The 4k7 can initially be set to its mid point.
The leds must have cathode, c, and anode, a, the right way round, otherwise serious damage may result when switched on after the transistors are added. It is a good idea to try connecting the power supply at this stage in the construction to check that the leds all have voltage drop around 1.7V to 1.8V, or just check that they all light up. For most leds the cathode is identified by either a flat on the body or by a shorter lead, but there are a few exceptions to this rule.
The capacitors are added in one of the next diagrams. The polarised electrolytics must be connected with the correct polarity, the negative terminal is usually indicated by a light coloured band down the side of the case and a row of minus signs. Take extra care when soldering the 10p ceramic, these are apparently easy to damage with excess heat, and this is a vital part of the high frequency stabilisation, without this the amplifier will almost certainly be unstable. Using some sort of heat-shunt clip close to the capacitor body when soldering is recommended, e.g. one of the clips sometimes used on test leads.
The last diagram shows the transistors. The 2SA1209 and 2SC2911 are shown with a thick black line to indicate the back metal part of the case. If a supply voltage much more than the recommended 60V is used it may be a good idea to attach a small heatsink to the 2SA1209 nearest to the top. This may not be essential, but will at least prolong its life.
The next diagram shows the insulated wire links used to reduce current loops in the output stage. Moderately heavy gauge stranded copper wire, e.g. the type used for speaker cables, is suitable. Adding the wires shown as red and blue first, the green wire can then be wrapped a few times round these to keep loops small. There is perhaps little benefit in twisting the wires tightly together, which could increase unwanted inductance. The external connections are shown in black, and the same sort of insulated wire can be used here. A central 'star ground' is shown at the top, and the wire paths should be similar to those in the diagram, for example keeping the supply leads close together, again to reduce current loops where the currents are highly nonlinear. The star ground can consist of several solder tags bolted together to the metal case at this point. Holes are included on the board for terminal pins at the input. The pins made by Vero for 1mm holes are suitable if required.
For a stereo amplifier this may not be the ideal arrangement. If the input earths are connected together at the input socket or volume control then there will be an earth loop because the two earths are also connected via the star ground. My experiments suggest that earth loops need not be a serious problem, it is the input signal loop we need to pay more attention to. Earth loops do not necessarily lead to voltages between different points round the loop even if a current is induced round the loop by varying magnetic fields (the current is produced by an electric field, but this is proportional to the rate of change of the vector potential, and the scalar potential can remain zero). If anyone wants to experiment with different earthing arrangements the next diagram is a possibility which can avoid the earth loop. I have tried both arrangements and found very little difference in practice, but in different locations there may be more external interference to worry about.
The earth lead from the input looks longer than is necessary, it could have been connected to the earth further up the board, but this is done to reduce earth and signal loops at the inputs. Only one of these connections is needed to one channel, the other channel input earth connects to this first earth as shown. This is in effect a second 'star ground' point for the input signal leads and amplifier input earths so that loops are minimised. There should be no further connection to the metal case at this point. A separate mains earth in the signal source should be avoided if this can be done safely, and most cd players etc. use double insulation power supplies to allow earth links to be omitted without danger.
Keeping everything simple is a good idea, and I prefer a direct connection from a 10k log volume control to amplifier input, and plug a single input lead from the volume control into the signal source. Pre-amps and input selectors are best avoided, unless the pre-amp solves some impedance or signal level matching problem. If an input selector switch is really necessary one possible improvement is switching both signal and earth inputs so that there are not a whole array of input earths connected at the same time, each with possible loop problems.