This one has been a long time on the work bench. This project started in 2012 but got side tracked and was only completed in April 2015. During this time the project changed somewhat, and instead of the Semisouth Jfet version which was first announced by Nelson on the Diyaudio forum, it ended up as the Diyaudio version which uses IRFP240 Mosfets in place of the Semisouths. The Semisouths are still in the bit box waiting for a suitable project.
It was decided to build this circuit from scratch but for the less confident it is understood that a pcb for this amp will shortly be available from the diyaudio site. The matrix board that was originally used for the F2 project seems to have become hard to find so it was decided to build this circuit on standard paxolin board. This caused a slight variation in the build process as the paxolin obviously needs holes drilling for the components.
The method used was to draw the circuit out on a sheet of paper with as few overlaps in the tracks as possible. This drawing was then transferred to a piece of paper the same size as the paxolin board, drawing all the components actual size. The drill holes can then be marked on this piece of paper. If the tracks are actually drawn on this paper then they can be used to bend the copper wire correctly for soldering underneath the board. This paper can then be fixed to the paxolin board with sticky tape and used to drill the holes in the correct locations. The images show the actual drawings used for the circuit boards.
The circuit boards have been attached to the heatsinks with 15mm stand-offs, fastened to the heatsink and the board with m3 machine screws. The heat sinks are the same as used in the other firstwatt projects, i.e 300mmx150mm from Conrad in Australia, specified at 0.25 degrees per watt.
The boards were laid out to allow the mosfets to be fixed near the bottom of the heat sinks(heat rises!), and spaced out across the heat sink to spread/share the heat transfer. As was mentioned earlier the jensen transformers were located toward the back of the case, as far from the mains transformer as possible. The 15mm stand-offs were chosen as they proved to be the matching size for the legs of the mosfets when they were attached to the heat sink.
The mains transformer was 2x18v 300va obtained from airlink, with all shielding specified on the order. Experience with the F2 build had shown that these toroids do emit an amount of hash so it was deemed prudent to ask for as much shielding as possible.
The chassis for this amplifier is one piece of 2mm aluminium sheet which has been bent to suit the dimensions of the heat sinks. This forms the bottom and the back of the chassis, to which all parts are connected. The chassis is connected to the heat sinks using 15mm equal angle 3mm thick and fixed using 3mm machine screws and nuts. The bottom has been liberally drilled with 6mm holes for ventilation. The angle fixing the top to the heat sinks has been tapped for m3 screws as it is the last panel to be attached. The top is 2mm perforated aluminium sheet and the front is a piece of 5mm aluminium.
The plot approximates to 0.066% distortion. The signal generator used was a diy construct detailed here, which itself has a thd no better than 0.025% so a better result is unlikely with the equipment at hand. The signal was captured using an Acer I5 laptop using a Steinberg UR22 digital interface. The method used to arrive at this figure is documented here.
This configuration seems very susceptible to mains fluctuations. In the first version of this amplifier a random selection of irfp240s were used from the bit box, attached to the heat sink with ‘sil-pads’. This version showed the bias moving quite considerably over a period of time, and was very difficult to get a stable idle point. It was first thought this may be caused by temperature fluctuations so a new set of matched irfp240s were substituted in, and the sil-pads were replaced with mica and goop. The new version displayed the same characteristics so more investigation resulted in these measurements.
It seems clear from the results that the mains variations during the test were quite considerable and had a profound effect on the bias point in this amplifier. With patience it was possible to set a bias point where both channels tracked together, and a point was set roughly in the middle of the high and low mains voltage positions.
More time is needed to determine if the mains fluctuations have an effect on sound quality.
It’s been running while connected to various different loudspeaker types: full range, two way, reflex, TQWP, transmission line and OB. So far it’s shone with all these types and it’s as enjoyable as any of the amps documented here. Intense scrutiny has found it slightly shy at the top end when trying to drive full range electrostatics, so this may be a combination to check first before committing.