Musical Fidelity A1 - Technical:

This page deals with technical aspects of the amplifier - construction, circuit details, faults and maintenance, etc. Before taking the amplifier apart, read the following disclaimer:

Disclaimer:

Construction and disassembly:

The top cover and front panel are extruded aluminium alloy, and the bottom, sides and back panels are painted steel. The top and side panels overhang the rear by about an inch, which neatly hide the connectors from view - I think that is a nice detail but I can recall at least one review criticising it.

Interior access is not easy. These units are not designed to be worked on! The side panels are easily removed, but you have to remove the top cover to access the main PCB. As the top cover is the heatsink, the amplifier can't be run for more than a few seconds without some other form of cooling, such as a fan.

The top panel is help on by 5 screws and the front panel (which is easily removed by undoing 3 screws underneath the unit). As you remove the top, you'll find that the heat-transfer compound will be gripping the aluminium U-section (which holds the output devices) quite effectively and some care is needed to separate them.

When I first took mine apart (2 days after purchase ;-), I found that a heavy streak had been applied down the length of the U-section which only covered 10-20% of the area - this would result in poor heat transfer. Perhaps this lack of quality-control at the assembly-line is responsible for some failures?

You'll want to clean up the heatsink compound straight away before it gets on your clothes and bench! Having removed the top, you'll see the PCB...

Note that this picture shows my amplifier after I had serviced it (apart from changing the four main power supply capacitors). In the process, I tidied up the internal wiring slightly - don't be surprised if you find a birds nest inside yours!

Early MKI amplifiers have a two-part top panel, as you can see in this picture, supplied by Frodo Muijzer. This is actually a picture of an A1-X, but as noted previously, this appears to be the European designation. Note that this model has ventilation holes in the side panels, suggesting that it is not a very early example. If you have a two-part top cover, note that the two halves need a layer of heat sink compound at their mating surfaces.

The smoothing capacitors are axial and hard to find. A quick look at the Farnell catalogue suggests that 345-0715 from the Nichicon VX series might be ok, but check the dimensions carefully before ordering.

To remove the PCB, you'll need to remove two spacers towards the rear of the case and a number of M3 screws holding the PCB to the bottom panel. Next, free the rear panel from the bottom and the PCB-mount phono sockets. Unbolt the 4mm phono earth connection. Finally, you'll need to free up the volume and input-select controls. Carefully remove the volume knob by rotating and pulling against an end-stop (it's a friction fit) and unbolt the potentiometer which is on flying leads. Then undo the front screw in the shaft-coupler and pull the input-select knob complete with extension shaft through the sub-front panel...

The PCB will now lift out, but pay special attention to the wires from the mains transformer, the speaker connections and the power LED. Told you it wasn't easy!

Close-up of the left output stage, showing re-labelled power devices... At this point, I'd replaced all the electrolytics except the main PSU reservoirs.	Close-up of the thermal cut-out that's in series with the mains supply. Note those bulging smoothing capacitors...	Underneath the PCB, showing the hardwired power supply rails and the loudspeaker outputs. Note the darkened etch-resist on the high-current paths.

As you can see, some good quality components are evident. Most of the resistors are 1% metal film. However, all the electrolytics are only 85°C rated, which surprised me!

Circuit Description:

While servicing the amplifier some years ago, I took the opportunity to sketch out the schematic. You can download a scan of the hand-drawn circuit here - be warned this is a large file (339KB). Use a graphics package to print it full-page - it should be reasonably clear, but I am working producing proper electronic versions - honest!

Power Supply:

The main power supply is relatively conventional - the mains supply is presented to the torroidal transformer via a fuse, a thermal cut-out and the power switch. A device that looks like a VDR is mounted on the power switch...

The secondary windings produce ±24V nominally using discrete diodes and 2 reservoir capacitors. The second pair of capacitors is fed via 0.47Ω resistors which reduce ripple by a good amount - I measured 0.8V pk-pk across C3 and 0.25V pk-pk across C4. A cheaper version of the choke PSU's that MF use these days...

The smoothing capacitors are rated at 25V - rather close to the limit in my opinion!

Power amp:

This is a curious design, in my experience at least. It is a symmetrical design, using what appears to be two identical amplifiers - indeed there's separate feedback paths for each. I guess that you could replace one half of the amplifier with a current source and it would work happily...

Several people have emailed to inform me that Tim de Paravicini designed the circuit - he is more famous for his valve designs, particularly single-ended, I'm told...

Each half is very simple - essentially a long-tailed pair feeding the output device which is connected in a common-emitter configuration. The two intermediate transistors (TR7/8 and TR5/6) are simply emitter-followers, providing current gain...

At a glance, it doesn't look like a promising circuit. The LTP is unbalanced and not fed from a current source (although the tail resistor is fed from a zener-regulated rail, which is better than nothing, I suppose). There is no local-feedback around the VAS (the output devices) - a small resistor in the emitter of each output device could have been employed, with a small headroom (and heat) penalty. Also, the output inductor has been omitted, and the amplifier relies on the 0.22Ω resistors for stability at the cost of damping factor.

I'm guessing, but I wonder if the symmetrical nature of the amplifier is what helps to overcome the non-linearities in the simple halves of the circuit. An alternative view is that perhaps these non-linearities are actually what form the 'sound' of the amplifier... I'd welcome comments and further explanation from anyone who is more familiar with this type of circuit - I'm afraid I'm strictly a 'Lin'-man until Douglas Self does a book on symmetrical amplifiers!

What sets the quiescent current? Well, I think you need to look to R30/31 for the answer. There is a small voltage across these resistors, and assuming that R30/R31 are accurate in value, you can determine that the standing current is 0.69A, resulting in 7.6W of Class A. Two factors influence this. Firstly, each half will have a small dc offset, due to the unbalanced LTP and other circuit conditions. In production, this is bound to be rather variable from unit to unit, so the 3M7 resistors (R6/11) appear to have been added to introduce a defined dc offset in the LTP, presumably swamping the other (poorly-defined) offsets. This will generate the voltage across R30/31, and the resulting standing current.

The output devices are standard 2n3055/MJ2955 pairs. However, Musical Fidelity labeled them in an attempt to make the amplifier appear more "mysterious", to presumably to ensure that their service department had to repair them (or sell transistors to repairers). They were re-labeled to "A1N" and "A1P" respectively. Also, a 4-digit date code was added - this confuses people because they assume it is part of the type number. Referring to my examples, my date codes are "9042" and "9034" - that means week number 42 and 34 of 1990. This form of date-coding is pretty-much industry-standard.

From what I've been told, the output devices aren't critical, and there are no special requirements. However, if you know any different, perhaps you could let me know. I know of people who use the Motorola (On-Semi) MJ15003 and MJ15004 pair as replacements - these are more highly rated than the originals, and should be more reliable in the long term.

A quick note about the power supply arrangements. R13/14 drop the unregulated rails down to the Zener-regulated ±12V rails. These are separate for each channel, but the line preamp is based around a quad op-amp which is powered from these rails on the right power-amp. Hence, the left channel has higher values for R13/14 because of the reduced current demand. On power-off, my amp fades to the left channel as the sound dies, and I think this is why...

A note about class A operation:

The claim of class A operation has always been a controversial matter. It's worth pointing out the Musical Fidelity say that the amplifier is "strongly biased into class A" - apart from the lettering on the front panel (!), they have never claimed that the amplifier is a pure class A design as far as I can establish.

In truth the A1 is, like most other power amplifiers, a class AB push-pull design. The quiescent current is simply set high enough to ensure class A operation up to a certain power level with a particular load impedance - around 8 watts into 8 ohms. Outside of this, the amplifier reverts to class AB. This is absolutely the right way to do it, otherwise low impedance speakers will cause clipping at artificially low levels. Classic Krell amplifiers like the KSA-50 will have enough standing current to provide 50W of class A operation to 8 ohm speakers (around 1.75A), however when driving 2 ohm speakers, it will revert to class AB above 12 watts. The prodigious low-impedance capability of this amp meant that it was frequently used to drive Apogee ribbon speakers...

Subscribers to audio newsgroups may have come across an Australian character who vehemently attacks Musical Fidelity at every opportunity, and while I perhaps might understand some of his sentiments, I wouldn't choose to be quite so polarised in my views. Anyway, you'll find that he claims that this amplifier only delivers the first 1 or 2 watts in class A, and judging by the e-mails I get, this view worries people. Rest assured that it is simply incorrect - allow me to present two arguments and a quote to support this.

Firstly, I found a several posts by him on this subject where he incorrectly calculated the class A power from the quiescent current, and in fact if you repeat his error with a standing current of 0.69A, you get 1.9 watts!

But, more objectively, let's think about the thermal issues. Apparently he's found the quiescent current to be of the order of 200-350mA, although no details of the measurements were given. If true, the output stage would only dissipate 20-35W instead of the 70W that mine does. I've measured mine as detailed above (and also measured the AC power draw to confirm this), and I've also measured the temperature rise of the heat sink. Based on this, I estimate the thermal resistance (heat sink to ambient) of the heat sink to be around 0.57 °C/Watt (40°C/70W). Now, if the output stage only dissipates 20 (or 35) watts, the case temperature would only rise by 11 (or 20) degrees C. As any A1 owner will tell you, they run too hot to touch, which implies a surface temperature of at least 60°C, which is a temperature rise of 35-40°C!

Finally, I'd like to quote from a Hi-Fi News review by Dave Berriman, published in the January 1994 issue. This review was admittedly of a mark 3 model (see below), but it's useful information anyway:

Out of curiosity, we checked the quiescent current through the output transistors. At 0.71A, this is enough for up to about 8W of pure Class A, which implies that over 8W the A1 runs into Class A/B.

I can think of a possible explanation for this - I wonder if models destined for hot climates like Australia were different to European examples? Within a week of posting this, I received an email suggesting that doesn't seem to be the case, but I'd love to explore this further minus the "Usenet vitriol". Please get in touch if you have some experience of these amplifiers "down under".

Line preamp:

And so, from an interesting power amp, we come to an appalling preamp, in my opinion of course. Look at the volume control - a total no-no! Fine for a preset adjustment that never gets moved, but for a user-control?

The volume controls, and input switches for that matter, get very noisy very quickly in the A1. I know someone who paid an enormous sum to get his pot replaced only for it to get noisy again after a couple of months. If you are tempted to replace yours - don't bother!

Despite what he was told, it has nothing to do with the heat. My 'scope runs just as hot, but has 25 year-old standard-quality pots in it. The problem is dc across the wiper, caused by the design of the preamp. Proof needed? Look at the PCB - holes are made for a small cheap pot (less than a pound from Maplin), yet the expensive (approx £10 from Farnell) Alps 'Blue Velvet' pot is fitted using flying leads. Recently, I serviced a MF B200 for a friend at work. This was a more expensive class AB MOSFET amplifier built in an identical case, yet despite the higher cost, it used the cheap £1 pot. Needless to say, this was a conventional arrangement with no DC on the track...

The other problem with this arrangement is the Johnson noise caused by the high resistances seen by the two op-amp sections. Note also there is no ultrasonic filter on the input, which could cause compatibility problems with some CD players...

The supply for this stage is taken from the ±12V rails of the right power amp, as discussed above. They are fed via the active filters T17/18, which take a few seconds to come up. Ever switched on the amp with a source playing, and wondered why you get a burst of music (always at much the same level), then silence, then the music at the expected level? When first powered up, the power-amps work almost instantly, but the op-amp is not yet powered. So, the audio signal travels straight via R33, R32, R34 and the volume pot. As the rails start to rise, the op-amp wakes up and the signal is muted until the rails become further established... That's my theory, at least!

Phono preamp:

Remember when integrated amplifiers had phono stages? This one even has an MC setting. It's slightly unusual in that it has a transistor pre-amp prior to the op-amp, and this provides the extra gain required for the MC option. It's quite neat - selecting MC (via the rear-panel switch) increases the gain by reducing the emitter resistance (at AC), simultaneously reducing the input impedance to provide the correct loading for the cartridge.

I'm not sure if I'd choose to use this configuration. For one, the input is open-loop, relying on local-feedback in the emitter of TR16 for linearisation. Which is much-reduced in the MC mode, of course... Also, it's a transconductance stage (voltage in, current out). I'm not sure how well-defined this is in this context. I'm also not exactly sure what TR15 is doing - at a glance it looks like current-source loading for TR16. Looking more carefully, it looks like TR15 is a voltage-source, but with the addition of C30 which appears to convert the stage into a pseudo current source. Any comments?

Phono PSU:

The power supply is slightly more involved...

I'm not sure why they've chosen this rather complicated arrangement when 78L12 regulators are so cheap. I wonder if they'd bought a job-lot of TL084 quad op-amps?! Or, perhaps it was so that their marketing dept could use trendy phrases like "shunt regulation"...

TR11/12 and TR13/14 are current sources, feeding the quad op-amp. This device is able to influence the supply rails by sinking/sourcing current via its output stages - the diodes ensure that IC1b only affects the positive rail and IC1c can only control the negative rail.

The surrounding circuitry is rather like you'd find in 'discrete' supplies - for example linear bench power supplies. ZD3 forms a reference voltage, filtered by R46 and C20, which is multiplied by the ratio of R49/R47 to determine the positive rail. As R44 and R46 are the same value and the negative reference is ground, the negative rail will accurately mirror the positive rail. All clever enough, but why?

Mark III Models:

Thanks to Nicolas Hodges, I've recently had the opportunity to examine a Mk 3 example. It is broadly similar to the earlier models, with the following changes:

• Power supply: extra smoothing capacitors and improved seperation between the two channels.
• Power amplifier: minor component value changes.
• Line preamp: a crude power-on mute circuit added.
• Phono preamp: Unchanged.
• Phone preamp power supply: the discrete arrangement has been dropped in favour of 78L12/19L12 regulator IC's
• PCB: lots of minor changes - the power supply and grounding arrangements have been improved, and the amount of internal wiring is reduced slightly. Interestingly, the PCB is marked "DAVID II" (see later)

Soon I'll post full details and pictures of this.

Final Edition:

The Final Edition is substantially different... it's actually a B200, which is a class AB MOS-FET based amplifier. As far as I can tell, Musical Fidelity have two basic amplifier topologies - the bipolar design described above, and the opamp-plus-MOSFET topology that was used in most of their expensive models.

Please note - this schematic is based on a B200 that I serviced some years ago, and there are some slight differences to the version used in the A1. Also, there are some errors in the schematic, especially concerning TR5 and TR6, so use with caution!

The input signal from the tape monitor switch is applied directly to the volume control in a conventional manner. This is the first plus-point here, because it fixes the problem with the original preamp - noisy controls are very rare with this circuit.

From there, the input is applied to an LM318 op-amp, an elderly, but respectable high-speed device. This stage amplifies and inverts the signal, but the voltage swing at the output of the op-amp is limited by the power supply rails. To develop enough voltage swing to drive the loudspeakers, the output stage requires some gain. This is provided by TR6 and TR6, and the MOS-FET's themselves.

As I said above, I've made some errors here so I won't attempt to explain this section in detail. I've been able to compare this diagram to some close-up photos that have been kindly submitted, but I need to examine an example "in the flesh" before I'd be confident in posting fully. But, TR5 and TR6 are actually arranged as common-base amplifiers, which provides some of the gain and the necessary level-shifting. The MOS-FETs are arranged as inverting common-source amplifiers, which provides more voltage gain.

The supply to the op-amp is provide by 13mA current sources TR1, 2 and TR3, 4. ZD1 and ZD2 regulate the voltage rails to 12V. Earlier amplifiers using this layout used simple resistors instead of current sources, but these are a welcome addition. I recently received an email from someone who was successfully running his European B200 in the US where the mains supply voltage is only 120V. VR1 is used to adjust the quiescent current.

You'll note that there are a good number of small capacitors dotted around for HF compensation. Stabilising amplifiers with such unusual gain structures is not trivial, so I would advise caution when considering any modifications.

From memory, I the B200 I serviced had an exciting, powerful sound quality. It certainly plays louder than the A1, as it was rated at 60 watts per channel. The 40WPC A1 FE has lower power supply rails and a higher quiescent current. I'm not sure how "accurate" it is in absolute terms, it sounded coloured next to the A1, hardly a neutral amplifier! One criticism - the signal to noise ratio of the B200 wasn't that good, so if you plan to use it at low levels with efficient speakers, you might find noise and hiss to be a problem. Later models used a modification which is derived from a National Semiconductor application note, whereby an external long-tail pair is used and connected to the compensation pins of the LM318. This reduces the noise of the amplifier.

What goes wrong?

Noisy pots and switches:

The bane of my life! It's so embarrassing to have a half-decent hi-fi that you can't adjust the volume! The failure mode is deposits on the track, particularly at the lower end of the scale. Due to the "design" of the preamp, the problem is compounded: as the wiper hits dirt and goes high-resistance, the volume is effectively max'd for a brief time. If the configuration was conventional, the sound would do the opposite, making the effect much less objectionable...

Also, the switches are affected for the same reason - the input op-amp stage puts dc currents through the switches...

As noted above, replacing the pot and switches is a short-term solution. If you retain the original preamp, the new components will go the same way.

Faulty electrolytic capacitors:

This is the second big problem with these amplifiers. Heat is a major enemy to electrolytic capacitors, so it will be no surprise to hear that they are a particular problem with this amplifier! The original components were fairly low quality to begin with, which clearly doesn't help.

Look in an electronics component catalogue, and you'll see that electrolytic capacitors have a finite life expectancy, typically around 2000 hours for an average-quality device. This is quoted at an operating temperature of 85°C, and as a rule of thumb, reducing the operating temperature by 10 degrees will double the life expectancy. As the surface of the heat sink is around 60-65°C, the internal temperature will be even higher. Remember, 2000 hours might sound like a lot, but that's only 2 years if you use the amplifier for 3 hours a day!

The symptoms of faulty capacitors are hard to predict. Basically, it only takes one degraded capacitor to affect the performance of the circuit, and as it's hard to detect a gradual change in sound quality, you can be forgiven for not noticing faults developing. When they get particularly bad, it's possible that other components can fail as a result of the changes operating conditions within the amplifier. The message is simple - if you have a fault with the amplifier, don't start trying to apply logic until you've replaced all the capacitors. In 95% of the cases I've come across, this will clear the fault.

Better quality capacitors will be rated at 105°C, and will have a longer life span as well. Follow the logic of doubling the life span with every 10°C fall in temperature, and you'll soon see why fitting anything less is a false economy.

Failed output stages:

The output devices expire with the heat, possibly prematurely if the poor application of heat sink compound that I discovered is anything to go by...

Luckily, this amplifier is relatively "fail safe", and not prone to the catastrophic failures that affect many audio power amplifiers - failure of an output device normally leads to cascaded faults further back in the circuitry, which is one reason why service centres are often reluctant to take on such jobs. But having said that, you should carefully check every active and passive component in the power amplifier stage before re-applying power. And make sure your speakers are not connected!

As there is no dc protection circuitry, a faulty output stage will probably damage loudspeakers if there is significant dc offset present... Maplin sell (sold?) a Vellmen kit that protects the speakers from the amp under these circumstances, and is definitely worth adding, if it is still available...

Maintenance:

Replace all electrolytics:

Discussed above, but so important that it's worth mentioning again. This is well-worth doing if you are going to the trouble of dismantling the amplifier. As noted, the heat inside the amplifier will do a good job of drying out the electrolyte, which could cause a degradation in sound quality and possibly other, more serious faults.

Remember to look for long-life, low-ESR, 105°C rated capacitors - Rubycon YXF, Elna RJH or Panasonic FC 's are all good starting points. I'd steer clear of "exotic" or "boutique" components like Black Gates as they're only rated at 85°C - it seems perverse to spend a fortune on capacitors that will only last for a couple of years.

It took me a little while to locate the main reservoir capacitors because of the limited space - they can only be 25mm high. RS stock them - part number 106-132, but unfortunately, they're only rated at 85°C. If anyone can recommend a suitable 10,000uF, 25V 105°C device that is readily available from the usual supplies, I'd love to hear from you.

If you like, you can replace C6 and C7 with 1µ non-electrolytics, but they have to be physically small. This is highly recommended if you build my replacement preamplifer...

Needless to say, you need to be very careful about the polarity of all the electrolytic capacitors. There are markings on the PCB, and you can check against the schematic...

Reapply the heatsink compound:

Which you obviously have to do if you remove the lid. So, if you fancy having a nose "under-the-hood" of your amplifier, but aren't brave enough to do anything more drastic, this is one which might improve long-term reliability...

This picture shows you how best to reapply the compound before reassembly. Obviously, all the old compound has already been removed from the U-section and the top panel. Using a straight edge, the compound has been spread evenly in a thin layer, avoiding the screw holes and the outer few mm. When reassembled, a small amount of compound is seen to escape from between the joint, indicating a good distribution...

Another point to check is the torque of the screws holding the output devices. Don't overtighten them, but just make sure they aren't loose.

After re-assembly, you might want to check the torque of the tree U-section bolts once the amplifier has warmed up. Don't apply masses of pressure, as you might strip the threads in the aluminium U-section, but just make sure they are 'nipped-up' securely...

Remember that if your top cover is in two sections, you must apply compound where the two halves meet!

Onto the next section - mods...