Simple Killer Amp!

[amplifierguru]

Conventional power amplifiers commonly feature a low voltage gain input differential stage followed by a very high voltage gain stage then usually a unity gain emitter or source follower. This design suffers from a distinct topological deficiency that degrades the sound and is responsible for inferior sound quality.

The typical gain distribution is shown in the thumbnail where the weakness of PSRR of the input stage ( not amplifier) due to the divider with stage output Z vs load Z determines intrusion of Class AB PS artefacts which are magnificently amplified by the very high gain Vas to output, blurring the stereo image detail and spatial info and raising the dependance for performance on the PS quality!

Better sound quality from such a design centres around fiddling at the edges of this flawed topology rather than adopting a new one.

[amplifierguru]

Ignoring the 1K input impedance glitch in the hastily prepared drawing, feedback theory will dictate that this level of intrusion or sensitivity is -

Vs out proportional to A2 x A3/A1 and also to stage load Z/collector Z.

So to reduce this harsh power supply induced output commutation distortion from degrading the sound we need to -

1. Increase A1 without increasing load Z proportionally so CMs are out!
2. decrease A2A3 so the intrusion is amplified less.

So ideally most of the voltage gain should be in the input stage and not in an after injection stage and especially not in a following high gain miller comp stage which degrades load Z as well and results in the typical thumbnail response.

So far so good

[AKSA]

But Greg.....

Your input stage here is still a diff amp!!

The conventional LTP is a voltage in/current out block; a transconductance configuration. It is the current output which normally drives the voltage amplifier. I find myself agreeing with some of your comments, but your block diagram still does not obviate use of a long tail pair.

Cheers,

Hugh

[amplifierguru]

Hi Hugh,

We can't throw out the baby with the bathwater here - diff'l amps have many advantages , not the least being good linearity and low temp sernsitivity in comlementary topologies.

Just about stumps for me till tomorrow.

Cheers,
Greg

[ThorstenL]

Konnichiwa,

Originally posted by amplifierguru
We can't throw out the baby with the bathwater here

True, but "diff amps" are the bathwater. They are about 50% of the problem of the generic Amp topology (the rest is the VAS and teh Output stage).

Why not go for complementary folded cascode current feedback topology.

You can get huge amounts of gain in effectively one stage, no common mode induced distortion, you can use good quality complementary J-Fets (2SJ74/2SK170) in the input and if you wanted follow the (single) Input/VAS stage with a BJT Buffered (Emitter follower) Mosfet Output stage

Sayonara

[ThorstenL]

Konnichiwa,

A picture speaks 1000 words, here a piccie of what I have in mind, principle diagram.

Sayonara

[darkfenriz]

I see that positive feedback gets very fashionable these days.



sorry for joking.

[MikeB]

That's a good one ! (but darkfenriz was faster )
You might have overlooked that fc-topology is inverting, means you applied
a positive feedback, your circuit is not functional at all...

Mike

[pinkmouse]

Thread for Jam's new topology is here.

[ThorstenL]

Konnichiwa,

Originally posted by darkfenriz
I see that positive feedback gets very fashionable these days.

Originally posted by MikeB
That's a good one ! (but darkfenriz was faster )
You might have overlooked that fc-topology is inverting, means you applied a positive feedback, your circuit is not functional at all...

Well, not so much really overlooked, it is a result of using "cut & paste" in a graphics programme to get a general indication.

Simply remove the NFB and keep the amplifier open loop, change the feedback application point to the common gates OR change the output stage from common drain to common source for one more inversion, probably a good idea anyway to swap common drain for common source.

Anyway, attached with changed feedback arrangement, you can still use easily 50K...100K input resistors (according to Borberly's work) and then scale the feedback resistor to suit.