Hi Os!
All in all I deeply appreciated the knowledge how to design good sounding gnfb amp, this was my grail for some time. True. In other words how to spoil the signal in high gain 10Hz open loop amps and then cure it with sophisticated gnfb. The solutions founded here opened my eyes and I started to believe that this is possible, I will try your proposition.
But like always, the barbarian wins, try to beat the designs like this, http://www.diyaudio.com/forums/solid-state/183335-modified-follower-99-hf-transistors.html , impossible..
Have a careful second look Os, The schema is about the same as the circuit that Bob proposes only its elaborated. After the Diamond buffer a driver stage is introduced to further enhance gain and the input transistors of the diamond are darlingtons. These are the only diffirences but the basics of the circuit is exactly the same. This was done to further increase gain but not only, I suspect it will increase the openloop bandwith as well. The pioneer uses this outputstage without feedback.
Same deal in the dynamic rail model of the early '80s.
(Technics SE-A1 competition, though Glen-K would likely have disapproved)
Last edited:
"Without feedback" ...at least that is different.
The extra stage and the darlington's must make for 1M current gain - WOW. The CCS "mute" with the Jfets is quite unique as well. Still , Pioneer threw a lot of parts at what should be a much simpler circuit.OS
Ostripper`s implementation of the triple is technically very nice indeed, however, the single current stage is much more "perfect". Now we`ve got five nonlinear base-emitter and base-collector junctions, serious phase shifts, poor bandwidth and degraded switching performance, Pawel may also worry about the strong increase in memory distortion; a good many monotonic and random distortions to iron out (actually to live with).
#1 -It survives the abuse of the hot Tennessee summer.
#2- with the led CCS's , it is simple (well , not as simple as a EF2)
#3 - using the output pair for drivers , feel free to pile up 10 pair OP devices and run it from 90V rails. A true high current OP stage - go ahead and make yer' 1000 watt amps - fools.
#4 - Really hear what goodness your VAS has in store for you as it is not under load. Feel free to use 3-4 ma low Ic VAS's (to-92 - mpsa92/42) , resistive shunts will remain static , irregardless of output load.
#5 - It does not blow up when pushed to the rails (clipped horribly) ... fully passed the electrical abuse test with a power hungry dayton 4R subwoofer.
It's better than TMC.
OS
Go with a schema like the Pioneer then
, it addresses many of the problems you mention, used in open loop though.Can you show us a single current stage that can do 200w up.
Can you show a schematic of what you have in mind.
homemodder,
when describing circuit topologies, for the sake of credibility, it is important to give a detailed account of what is lost, not just what is gained, otherwise the arguments will sound like coming from a marketing brochure. Nothing can be gained without giving up something. That something is usually linearity.
The common collector amplifier is not as linear as generally assumed. For instance, the very high internal feedback results in negligible isolation between input and output and together with an applied overall feedback it makes amplifier performance hugely depend on the load and signal processing become a mess. This is very far from the ideal conditions and perfection.
when describing circuit topologies, for the sake of credibility, it is important to give a detailed account of what is lost, not just what is gained, otherwise the arguments will sound like coming from a marketing brochure. Nothing can be gained without giving up something. That something is usually linearity.
The common collector amplifier is not as linear as generally assumed. For instance, the very high internal feedback results in negligible isolation between input and output and together with an applied overall feedback it makes amplifier performance hugely depend on the load and signal processing become a mess. This is very far from the ideal conditions and perfection.
My 3 pair contraption will do that. 5-8 pair (below) would do much more. Still 40uA VAS load to run that 170 p-p waveform (4R). I don't need that much - 3 pair and the dayton sub is shaking the whole house.
OS
Attachments
Hi WuYit,
Could you describe what you mean by the "single current stage"?
Do you mean a VAS directly driving a single output emitter follower pair?
Cheers,
Bob
I could imagine a fet-based "my-dream-amp", with a "single current stage", consisting of one "input+VAS" stage with 150V shunt PS rail (2N5278 j-FET), which drives the follower-like output stage based on radar-LDMOSfet, with 600W dissipation, Ciss near 100pF and near 10S transconductance at 5A idle current (can be chosen among NXP products).
Although not realistic for DIYers, but quite real and unbeatable sound-wise.
Last edited:
Hi WuYit,
Thanks for clarifying that. I think that keeping the number of stages to a minimum is a central MISGUIDED issue.
I'm not saying that we just throw transistors at a problem, but there is an optimum approach that should not be constrained by some one-size-fits-all philosophy.
Have you simulated the kind of design that you are talking about? Have you built and measured the kind of design you are talking about? When you express concern about nonlinearities of multiple stages adding up, you are talking about something that is readily simulatable or measurable.
Cheers,
Bob
With my deepest IMHO, I would note, that many people, who assembled and listened many amps in good systems, including Nelson Pass among them, don't agree with this statement.
Why? Simply we do not measure correctly nowdays. The multystage designs quickly hit into fundamental limitations, related with electromagnetic interactions of parts, wires and traces. Also we accumulate effects of parts imperfections, especially of caps, that are completely out of any simulation soft. Yes, we can hardly measure these effects, similarly like we can not measure effects of cables or binding posts.
At the same time, if one designs an amp for a subwoofer or for a bass section of tramping system (more power, more power ...), than all things I mentioned here can be forgotten. Just follow standard approach, or jump onto modern level, class D.
Last edited:
That' an interesting point I read from time to time.
Do you have any evidence that would help such an opinion? How can we speak about electromagnetic or interaction if nobody makes measurement (correct or not) to support this? If measurements or simulation can't bring any evidence, how can a listening test point out the precise point of dissatisfaction?
That would help me to clarify what to correct if I hear something strange I can't imagine how to correct on the basis of measurement and that appears only by listenig. Well I know my ears are not perfect either.
To add some points, EMC simulation softwares exists (we use them in my company), it is today possible to extract all parasitics and coupling from any pcb (or integrated circuit!), and any component can be measured to define its parameters up to a few GHz. I guess it would not be usefull to correct a subjective listening feeling.