Q1 Can I use a polypropylene or teflon input coupling capacitor with the MJR7, and increase the value for better bass?
A1 There is no reason not to use different types if you believe you could hear a difference, provided they are low leakage. I have never personally heard any difference between capacitors, cables, resistors, or any other passive components, but my hearing is probably not so great, so maybe there are some people who can hear much better. Conventional distortion measurements suggest that only a very few components would have anything remotely close to audibility, e.g. some high-k ceramic capacitors. For an input coupling capacitor the idea is to choose a value sufficiently high to reduce any signal loss, so given a correct choice the signal across the capacitor should be very small, and any nonlinearity therefore also very small. The 2u2 value reduces output by only 0.5dB at 20Hz, far below variations from even the best speaker in a well designed listening room, so increasing the value is probably pointless, and may increase the switch-on thump. An increase to 4u7 may not do too much harm. I am tempted to do a distortion extraction test, using a music signal, to compare cheap polyester and electrolytics to the more exotic varieties. I just saw one example now available from Farnell, a 'paper in oil' type 'for high performance audio', at £21.45 for 2u2. These are 95mm long, so a bad idea for input coupling applications where pickup of interference needs to be kept low.
There are many seriously nonlinear capacitors in every solid-state audio amplifier, these are the junction capacitances of the transistors. The base-emitter capacitance at high current is mostly the diffusion capacitance, and this is proportional to emitter current, so for example in a class-AB output stage this capacitance can easily vary by a factor of 10 or more over a cycle of the signal. In contrast a polyester input capacitor may typically have a variation less than one part in ten million. Using a non-inverting amplifier there can be significant input capacitance nonlinearity. There are ways to minimise these problems, for example bootstrapping the input devices so that their operating voltage is kept almost constant, or the method I used, to add a cascode stage and use a 'virtual earth' input.
Q2 Can I use different transistors? What about vertical mosfets, these are much cheeper.
A2 I am making a list of alternative transistors, but so far I have found nothing much better than those already specified. The mosfets must be lateral types because these have about zero temperature coefficient at the 100mA operating current. Vertical mosfets would need temperature compensation. Lateral mosfets also appear to be less easily destroyed, so the simple supply line fuse I used may be adequate protection.
For the Hitachi 2SC2547E and 2SA1085E in the input stage alternatives are Hitachi 2SC1775E and 2SA872E, and also the Toshiba 2SC2240BL and 2SA970BL. The noise figure of the NPN device is important, it needs to be low at 10k source impedance and Ic 0.5mA. To achieve this high current gain is important, and low rbb' unnecessary, though unusually the 2SC2547E has both high gain and low rbb'. The types listed are mostly around 0.5dB. The 2SA872 was the input transistor specified about 30 years ago in the original Hitachi power mosfet application note, and the 2SC2547/2SA1085 were available soon after. I bought these from a company called Ambit International back then (Ambit had a really useful magazine/catalogue called 'Tecknowledgey'). They are still available from Nikko Electronics although there are rumours that they are no longer being made by Hitachi. RS also have them but list the manufacturer as Magnatec.
Q3 Why no listening tests? How does the MJR7 sound compared to the xxxx.
A3 There is actually one listening test, it appears on the MJR6 page, and is far more sensitive than most conventional listening tests. The error signal was extracted when playing music through a speaker, then amplified and recorded. Anything unpleasant being added to the music, or anything being removed, would appear in this error signal and should be easy to identify.
I started making my own amplifiers over 40 years ago and have rarely had a reason to buy any commercial amplifier, so I have only my own previous designs to compare to. To my surprise I did hear what I thought was a big difference when I compared my MJR6 to an earlier mosfet design, and this was worrying until I remembered that the old amplifier is non-inverting while the MJR6 is inverting. With an asymmetric signal and a speaker adding significant even-order distortion there will be a real difference in the sound. Neither is necessarily more correct. Reversing the speaker leads on the MJR6 or MJR7 will reverse the inversion if required.
Q4 What happened to the 'feedforward output stage? Is there going to be a new version eventually?
A4 The reason I more or less abandoned that approach was explained in an article Power Field-Effect Transistor Amplifier. To summarise, the feedforward design needs heavy local feedback in the output stage for best performance, and also needs high current gain to ensure the error amplifier stays in class-A. This is more easily achieved, I thought, using power mosfets. Then I realised that if we use enough feedback to linearise a single mosfet in one sub-amplifier then we could equally well linearise a complete mosfet output stage with local feedback. A complete output stage is actually more linear than a single mosfet at the same quiescent current because gm is higher. Adding lots of local feedback is possibly easier than lots of overall feedback in a complete amplifier, but if the complete amplifier is sufficiently simple there may not be much difference. From this idea the original MJR6 evolved, followed by the improved MJR7.
If I wanted to make a bjt output stage I would still use the original feedforward version. It would be interesting to see how it works using the MJR7 input and driver stages, so I will not say 'never'. I did have one small improvement I wanted to try.
Q5 How can I get in touch?
A5 An address for enquiries concerning the MJR7 design is firstname.lastname@example.org but remove the two letters x and z (this is to prevent automatic spam emails, I already get about 100 per week at this address). Use AMPLIFIERS as the subject, otherwise it may get missed and deleted with the spam. I am sometimes slow to reply so please be patient.