Quick question on output stage

[akis]

Has anyone made or considered making an output stage just? Assuming you have an input singal of 50V rms at the end of a cable, would you be able to build a PSU + stage where you'd simply just modulate the DC rail to drive a set of speakers? So there would be no input stage, no VAS ?

[analog_sa]

It is an existing concept usually called a power follower.

[Itsmee]

Have a look at the F5 in the Pass forum

[AKSA]

Most cinema amps are split into two sections; input/VAS at the projection room, and power buffer near the speakers.

This protects S/N ratio over long lines.

Hugh

[akis]

I was reading a paper into amplifiers and it seems the majority of effort (dozens of pages with theory, design and patents) has been applied around :

(a) VAS swinging as close to rails as possible
(b) VAS needing a constant current source to achieve stupendous gains
(c) global negative feedback required to tame gain and distortion introduced by above
(d) input stage must be differential for negative feedback to be applicable
(e) input stage succeptible to supply rails fluctuations thereby needing complex regulators/mirrors
(f) negative feedback affected by phase shifts and dozens of ways to apply it in parts, trying to combat said shifts
(g) a few dozen pages dedicated to esoteric nuances that I now forget.

All the above seems like kludges to me, unless there is no other way. But is there no other way?

I was thinking:

1) Who cares about swinging VAS close to the rails. Use a separate PSU and up the rails so the VAS can drive the output stage to the max before clipping. A separate PSU for the VAS is so easy to make and does not require a small PhD on how to get it right, because you'd simply use a simple stage with local (degenerative) feedback, no capacitors, no phase shifts nothing to worry about.

2) No need for global feedback or for stages with huge "open loop" gain. You simply need to take a singal of 1V rms (or thereabouts) and multiply it to 20V-60 V rms (or thereabouts).

I was thinking perhaps 3-4 transistors for the "input/VAS" stage with its own PSU (transformer, bridge and caps) and an output stage with its own PSU and low demands for input current such that the input stage can comfortably drive it. The output stage would be common collector push-pull for example or even class A common collector, the input stage would not know or care.

Perhaps the extra PSU for the input stage would add to the overall cost (say 20 pounds), but with the gain of simplicity? A DIYer could even construct and house the stages separately as there is no link between them (except the signal).

What do you think?

[traderbam]

Originally Posted by akis

All the above seems like kludges to me, unless there is no other way. But is there no other way?

Kludges? Bear in mind that the idea here is to reproduce sound pressure waves of adequate fidelity to fool the human brain using bits of processed sand and metal, rubber and some chemicals. Kludges may be unavoidable.

I was thinking:
...
What do you think?

I am wondering what your goals are? That is, are you mainly focussing on architecture and component count? Are you concerned about ease of build/cost? Are you concerned about how well the circuit fools the brain?

[CBS240]

Hi

There is no rule that says you must have global FB. In one amp I built, the best by far, the first amplifier, or pre-amp, is a complete loop, but with only 2 stages and a CL gain of ~20. It is a complete amp within itself, but is directly coupled to the second amplifier circuit or final VAS, and output stage. The VAS, Av=~5, contains its own local loop/compensation as does the output stage (error correction). DC output is controlled by a servo with a global loop, but I don't consider this global FB because the low pass Ft is set ~3 Hz. I have no problem with GNF for DC or very slow signals.

Not only is it not complicated to have a second supply for the input and VAS, it is preferred. The relatively small current needed allows you to use voltage regulators with little penalty in wasted power. In the above mentioned circuit, the output stage is source follower hexfets, so a second supply is required for full output drive. The outputs and bootstrapped drivers run from the lower +/-24V, high current supply. The other supply is twice the voltage but only ~20mA is needed. Since I only need +/-41V for my circuit, I used simple half wave doublers (+/-48V) followed by a series-shunt regulator. With only 20mA, cheap small signal parts can be used and dropping a few volts doesn't cost that much heat. Simple solution, works great. I tried full wave doublers but I never could tell a difference in performance so I’d rather save the parts and PCB space. BTW, the extra voltage allows me to use cascode VAS.