1970s Design Indulgence

Graham Slee wrote: BAK wrote: Using a regulated supply to the IPS and VAS transistors was found in early 1970's amplifiers. This is very true. The above circuit however does not use a regulated VAS supply. The constant current source started to emerge around 1975 from what I can gather.

The 70s design indulgence is of epic proportions and it looks set to continue in that vein.

I think by now I've exhausted every avenue of this particular configuration since I started investigating it in August last year.

Perhaps now is the time to conclude this one and archive the results, keeping the amplifier as it is as a working example, and decide what to do next.

In the beginning of transformer-less transistor amplification was (as the reader will now be aware) the Lin circuit and this is what I've tried to develop into something better over the last 12 months.

The original Lin circuit can be seen here on page 174, fig 2: https://www.americanradiohistory.com/Archive-Electronics/50s/Electronics-1956-09.pdf

Simply stated, it comprised a bootstrapped VAS and quasi-complimentary output stage to which was added a PNP input stage. Similar circuits were manufactured by Leak, Mitsubishi and Dynaco to name but three.

The almost universally stated reason for the Lin quasi-complimentary configuration was because PNP transistors were quite poor by comparison with NPN’s, and only used where there was no other choice. But in reading the Lin article there is absolutely no mention of this, and all but one of the transistors are PNP! OK, they're germanium and not silicon.

The blame for the "transistor sound" has always been crossover distortion, with JLH saying class-A was always employed in conventional valve operated power output stages. My interpretation in reading about KT66 and KT88 valve stages is that class-A was NOT always employed in conventional valve operated power output stages!

My experiments lead me to think that the biggest contributor to "good sound" is the linearity of the voltage amplifier, and my latest experiment in using a constant current source for the VAS tends to bear this out.

The bootstrap fails by being unable to deliver sufficient current at high frequencies. The twofold problem is that at high output stage current, the current-amplification factor usually diminishes; and it also diminishes at high frequencies where beta droops on its approach to the transistor's transition frequency.

Take the situation of several lower frequencies in music happening at the same time as a singer is voicing something sibilant. Where the waveforms coincide you have both problems above happening together.

The bootstrap delivers what the output stage is trying to do; to in-turn drive the output stage. What comes first: the chicken or the egg? That's anybody’s guess, but it can be seen that chaos must ensue.

The current source suffers none of these ills except perhaps running out of current. If the output stage beta is insufficient then the current draw might be sufficient to worsen VAS linearity. NFB will be useless at correcting it, and so we need to make the output stage have as much beta as it possibly can.

We can make the drivers have more standing current or less. By making it less there isn't as much wasted VAS current at small signal levels, but the drivers are medium power transistors and so have their betas designed for a reasonable current spread.

I guess it stands to reason that if the drivers are set to the minimum standing current indicated in datasheet DC current gain graphs, then at least it should be in "the envelope of intended usage".

The inquisitive might ask "why don't you know these things?" The answer is quite simple: it's because there are few designers capable of making high performance amplifiers, and that's because it isn't taught. Think of it this way: if it was taught, the world would be full of high performance amplifiers - there'd be nothing but high performance amplifiers - high performance amplifiers would be a commodity. Instead, we have a world full of mediocre amplifiers which suit most people, and transistor manufacturers produce data for mass producers where "it will do" is good enough. "It will do" isn't good enough for me and a minority of the population who feel the same. The result is that a few design engineers are enthusiastic (or crazy) enough to try and figure-out where the "sweet spots" of every part of a design and its components are. It isn't taught - it's learnt.


Edited by Graham Slee - 11 Aug 2019 at 5:38pm