Briarsfield Hi-Fi - Akai M-8 restoration

As you saw from part 1, the amplifiers were to be converted into a fully working integrated amplifier - this would be no mean feat.

The design called for several audio line inputs as per a regular audio pre-amp, originally the only inputs were the microphone and line inputs so that part needed completely redesigning. Because of the location of the input switch, relay switching was decided on for several reasons – firstly because of the combined length of the audio cables routing to a regular switch, but also because space around the selector switch would be tight and good quality selector switches tend to be bulky, not to mention expensive.

By using relays to switch the audio inputs, the signal lines are kept as short as possible, if well designed there are substantially less issues with crosstalk or connection problems as there might be if inexpensive switches are used to switch the audio signals themselves, also it's easier to ground the unselected inputs to cut noise and crosstalk (normal switches would need extra sets of terminals to do just that). With this design, a simple, inexpensive selector switch can be used because all it will ever do is switch the +5v power supply feeding the relay coils and the indicator lamps on the front panel. If good quality signal relays are used, it will work as well as any high quality audio selector switch would. Inexpensive rotary switches are still reasonably rugged and still feel positive to use, there's no problem at all with using one.

Originally, I was going to build the circuit onto a regular strip board but I wasn't happy about building that amount of complexity without a regular PCB layout, especially where low level signals were involved – problems with crosstalk and noise could easily become a problem. Besides, once you build circuits onto the board they inevitably become a mess, with even the best intentions the board layout will most likely look awful by the time you've finished with drilled or split tracks and wire links all over the place. Besides, a professional finish means a lot to me.

I designed the boards in Linux using KiCad, an open source program which handles both the creation of schematics and also the actual laying out of the PCB itself. Even with a reasonable background in computer design, KiCad isn't easy to get the hang of but once you do, it's a solid program and it can create professional looking board layouts. The thing to remember about a program like KiCad is that it forces you to be organised, you can't just go in and start clicking around adding random components to create a board. In KiCad you need to create a schematic first and then assign all of those components to case styles and footprints before you even think about putting them onto the PCB – it's a lot of planning but it pays off when you end up with a professional looking board layout where all the components fit well. Really, this amount of planning only becomes worthwhile if you're planning a production run of boards, where the time spent planning everything out would reap rewards when you could easily duplicate the boards without errors. That said, it's still nice to have a professional looking finished product, even if it is only a one-off.

I developed the board the old tech way, using toner transfer paper to get the track design on and then etched in ferric chloride for an hour in a warm place; I am seriously considering investing in the equipment to etch UV sensitive boards, this is a much more exact approach for anything more than a basic prototype and you'd need that extra accuracy if you were attempting complex board layouts. The toner method takes some care and is no good if you are trying to create a large number of boards, but for single boards it can work well.

The etched and cleaned PCB, not drilled yet.

This was the first time I had attempted prototyping circuit boards in a long time and my first time designing them on computer, overall it went very well. With hindsight, I could have made the pads around some components larger and I accidentally omitted the shunt diodes on the relay coils (the result being a loud 'pop' sound through the speakers every time the inputs were switched, quite alarming until I realized my mistake) - these were easily added in after, neatly soldered in on the track side of the board. Apart from this, everything went well and the circuit worked as expected from the start.

To keep things as simple as possible, I decided to keep everything to do with the audio inputs/outputs and switching on the rear panel – this meant a rear panel containing all the sockets (mains, line audio, speaker) and attached to that the PCB which contained the switching relays and also the regulated 5v power supply to feed both them and the LED source indicators on the front panel.

The power supply was originally to be taken from a heater winding on the left channel's mains transformer, the transformer had ample rating and added to that the fact that the oscillator valve was now removed allowed more than enough power to be tapped from it. The 6.3v AC supply was originally to be taken from the transformer on the left channel amp and fed through a rectifier and +5v regulator circuit (relays and LEDs all rated for 5v), I eventually powered it from a small auxiliary transformer which also powered the VU meters.

Rear panel assembled. This was more of a prototype, afterwards I decided to change the design and made a new chassis from aluminium

Display

A display was to be fitted to show what input was selected, this came from a donor amplifier of some description. Originally using lamps, it contained 4 clear plastic panes with PHONO, TAPE, TUNER and CD/AUX engraved onto them so that the lettering lit up. High brightness LEDs worked well in place of lamps - these were run from the selector switch in parallel with the relays.

Input display. Red, white, green and amber LEDs were fitted and held in place with high tech putty otherwise known as Blu-Tak..

Chassis

The rear panel was originally to be made from some black coated steel sheet I had spare, I decided to change this for a full chassis because I wasn't happy with the overall look and the prospect of all the wiring being tugged about when it was all being fitted into the case.

I changed the design and made a new complete chassis from a combination of 1.2mm aluminium sheet and 1.5mm aluminium profiles, designing and building something like this from scratch with hand tools is no easy task but with care the finished product can be good. Smaller holes were drilled using engineering grade twist drills (buying a kit of good engineering drills gives you a LOT of sizes to choose from), the larger holes were punched for neatness - try drilling anything larger than 8mm through metal sheet and you're asking for trouble.

However, punches are only available to cut down to 10mm or so, hence there's no alternative but to drill the smaller holes manually. I planned the panel out in a 2D CAD program and then exported the plan to a regular graphics program to add the lettering and logos, these were then converted to negative, mirrored and printed onto transfer sheet and applied to the prepared metal panel using heat and lots of pressure. The large 'A' logo was the original circa 1960's Akai logo, copied from the nameplate on the front panel of the tape unit and reproduced as a vector image in Paint Shop. The Briarsfield logo was added at the request of my customer, it wasn't my idea.

The grille was added for obvious reasons, that was a steel mesh which was primed and sprayed anthracite - I bowed it slightly to make it fit tight against the rear panel so it wouldn't rattle.

Once cut and drilled, the panel was wet sanded to produce a brushed finish and once the lettering had been applied it was then lacquered, this lacquer then wet sanded until relatively smooth and finally polished, like a pseudo clear anodized finish.

The lettering was experimental, this was difficult to make work and although the result wasn't perfect, I was still quite pleased with it - silk screening would have been the ideal but that takes a large amount of money and investment in equipment, only worthwhile if you had the space and enough use for it. The problems I had here were from the heat necessary to make the print stick to the metal, the metal expanded slightly more than the printing and hence smudged the lettering in places. Really, this was a test to see whether I could get something approaching a professional finish using basic materials.

Rear section of chassis assembled. A totally huge amount of work to get this level of finish.

VU meters

The amplifiers originally had VU meters fitted, these were powered from the loudspeaker output and were each fitted with a pair of 6.3v lamps, which were powered by one of the valve heater windings on the transformer. The design called for the lamps on these to be fitted with a combined dimmer and on/off switch and for the meters themselves to be powered from the signal circuits in the pre-amp section of the amplifier instead of the loudspeaker output as they were originally. The dimmer circuit for the lamps was simple enough, the original lamps were all blackened with age so were replaced with miniature screw cap lamps, 6.3v/150ma each. The dimmer circuit was built from an LM317 adjustable voltage regulator, the circuit configured to provide 6.25v across the lamps when at maximum. This circuit also needed its own rectifier to provide a DC supply suitable for the dimmer module. The regulator itself was fitted to the metal chassis to help keep it cool (acting as a large, conveniently placed heatsink) – however, as the metal tab on the LM317 is connected to the voltage output, it needs to be electrically insulated from the chassis while still being connected thermally.

Driving the meters themselves from the existing audio circuitry was a challenge, under test they needed around 2.8v to drive them into the +3dB red band – no part of the existing audio circuitry could provide this amount of voltage swing. I decided to power them from dedicated op-amps, these would take a voltage from any point of the audio circuitry that I chose and amplify it enough to power the meter. The idea here being that an op-amp has a very high impedance input and hence copies the signal which you feed into it but doesn't place any load that signal, it has virtually no negative effect on the audio circuitry at all.

I tried several different circuits, in the end it became clear that I needed to build a complete miniature amplifier for each meter complete with gain control and power supply to make everything work well. The circuit was a non-inverting type op-amp with a maximum gain of 10, I chose non-inverting because I was concerned about any load being placed on the audio circuitry. The op-amp used was nothing special, merely an LM741 – easily good enough to drive the meters. The inputs and outputs were DC coupled and the power supply on each one included a rectifier and a pair of 330µF capacitors, the value of these being certainly more than enough to provide smooth, stable voltage rails to drive the meters.

Originally I was intending to build the circuits onto a custom designed PCB as I had with the input board. However, I built the test version onto tripad board and altered it several times; at that point it worked well, was the right size and looked relatively neat so I left it alone – designing, developing and cutting a custom PCB would have added a great deal of extra work for no gain.
I originally planned to power the VU's from the heater windings, in practice this caused some problems with noise and also the lack of voltage available and also the lack of a centre tap caused its own issues with DC offset and bias - the bias at least would be easy to correct but would still leave little voltage swing, the 6.3v from the heaters being enough in theory to push the meters into their red bands but in practice a small separate transformer with 2 x 6v windings being much better. The amplifier boards were fixed to the back of each VU meter. I fused both sides of the AC line for safety, again this is overkill from a design perspective but it doesn't hurt anything.

The design also called for a set of tone controls to be fitted. This proved a conundrum to me, after I had agreed to design them in and sat down to do the math, it was clear that things weren't going to add up together without some changes elsewhere. The problem with tone controls in valve equipment is that they're often passive and sit inline with the audio signal between two stages. Nothing wrong with this approach at all but inevitably you'll lose a large amount of gain as the controls can only provide a loss of volume level, at least if you were designing an active network you'd usually sit it in some kind of local feedback network and you'd add an extra valve or transistor stage, correcting any gain you would lose in the first place.

I built the circuits on part of the main tag board on each channel which had been cleared of components during the upgrade. As they would only ever see around a volt of signal, 1/4 watt metal film resistors and LCR polystyrene capacitors worked well. The actual tone pots aren't anything special, merely 250kohm Alpha guitar pots.

The results of the first test after the controls were fitted was predictable, what sound was there was good and the tone controls worked as expected but the volume level had been severely cut - after all, the tone circuits were consuming over 20dB right across the audio spectrum when at their neutral positions and that's a heck of a lot of signal voltage to lose - you're effectively dividing the input signal by 10. After reverse-engineering each stage of the ECC83 the old fashioned way with a loadline plotted on paper, it was clear from this that each stage was already giving a gain of between 60 and 70 and squeezing any more from them would be extremely difficult, especially on the existing 250v HT supply.

Instead, I had a lucky break - it seemed that the first ECC83 stage would be able to handle around 1.2v of input before saturating, by removing 20dB of cut from the attenuation networks for the line inputs, I was able to compensate but still keep the signal levels comfortably below this threshold. This wasn't the perfect solution and I realized that I was pushing the first ECC83 stage pretty hard regarding sonics, but the extra gain did help things a great deal. This change in design meant that the tone switch would need to be fitted with a -20dB attenuator on the tone defeat position to match the volume levels, but that was easy enough to modify.

Another good thing with simple valve amplifiers is that the final signal which goes to the EL84 amplifier valve could just as easily be run through a basic DC filter network to instead drive a separate power amplifier. This isn't the ideal solution, as valves like high impedance and most power amplifier inputs will load the preceding stage to somewhere around 50K ohm, fine for transistors but far too low for valves to work comfortably with. In practice it worked well, some slight compression towards the lower bass but the sound was good.

The internal amplifier is reasonable quality, compared to the average 'consumer grade' transistor amplifier, it does compare well at low listening levels, especially on vocal or acoustic music - good components and clean signal path help greatly with this but the amplifier still has a large part to play in the sound. Power outputs matter more to transistors than they do to valves, a 5 watt transistor amp would sound pretty pathetic even if built with good components - many valve supporters say that 1 watt of valve power equals at least 2 watts of transistor power and that the first watt is the part of the power spectrum which matters most.

However, our 5 watts still isn't a great deal of power, although single ended undoubtedly has a nice low/mid region for low to mid volumes and ours had a very nice low end despite the small output transformers. That power doesn't go very far and even though you won't notice hard clipping like you would with transistors, you'll soon notice the lower frequencies start to bottom out and lag and a hint of ringing as the volume level increases. Still, for close listening at normal low/mid volume levels they could handle a pair of 8 ohm floorstanders with aplomb, this is where they really shine.
I took the pre-outs from the final ECC83 stage, before the coupling capacitor to the EL84 output. Under test, this output would easily peak 1v at full volume, more than enough to drive the average power amplifier.

Originally, the EL84 power amp section was to be switchable, so they could be used as pre-amps driving an external power amplifier. This caused several problems, firstly working out a way to switch both HT supplies, both heaters and both audio grid inputs. Finding a suitable switch off-the-shelf for all these purposes would be near impossible, relays would be ideal but then there's the question of space (which was already starting to become limited). Also I was worried about possible problems in the event of one HT relay failing for some reason but leaving the other HT feed powered, this could cause some serious heat damage.

Also, the mains transformers are quite small and small transformers inherently have poor regulation. Once the current draw for the EL84 was removed, the HT voltages would inevitably shoot up dramatically - from around 260v to 370v - this could have been explosive if the original 350v power supply filter caps were used, luckily the new ones were rated at 450v and could handle this extra voltage, but it's worth considering. This extra HT voltage would also add more gain, all of this combined with the removal of the EL84's negative feedback network would add another 25dB of gain which would send volume levels far too high and cause very audible problems with microphony and compression between the two ECC83 stages.

Between the loudspeaker sockets was fitted a push switch to switch the loudspeakers on and off. With the switch in the 'off' position, the speaker outputs are disconnected and the amplifier is run through a dummy 8 ohm load - you need this to keep the correct frequency response if running as a pre-amp or using headphones. This switch was the original record/monitor switch, it was open chassis so could easily handle a couple of watts. It was fitted to an aluminium bracket and screwed to the rear panel with spacers.

Curiously, the pre-out filter capacitor is one of the largest components in the amplifier, even though it's relatively small in value. Even if I wanted to, I couldn't use an electrolytic here as we're running the pre-outs from the second ECC83 stage anode, which runs at around 125 volts relative to ground - the average audio grade electrolytic normally used for DC audio coupling would explode at these voltages, allowing the 125v anode voltage across the inputs of the connected power amplifier - something I assure you would not be a nice experience. This is a normal 2.2uF/400v polypropylene capacitor, virtually the same component that you would find inside a good pair of speakers, used here to filter out the DC and give us the low sub-20Hz cut-off frequency which we've tried to maintain through the rest of the amplifier.

This isn't ideal, usually it's advisable to keep impedances higher and use smaller capacitors as smaller capacitors are regarded as more perfect - alas, real life isn't this simple and as the majority of (transistor) power amplifiers are between 47k and 100k input impedance, we've no choice but to use a relatively large coupling capacitor in an effort to keep the low frequency roll-off below 20Hz.