Gunderson compensation..

[mikeks]

What do y'all think of Steven Gunderson's approach?

Note that this is the technique used by Rotel in their RB1090 power amp., and Randy Slone in his designs.

Also available as:

'A Topology to Linearize Miller-Effect Compensated Amplifiers'

JAES Volume 32 Number 6 pp. 430-434; June 1984
An additional topology to Miller-effect compensated amplifiers, which reduces high-frequency distortion, is described. The addition is implementable in discrete or integrated circuits, with little or no additional complexity required in designs incorporating short-circuit protection. This topology is effective in reducing distortion to frequencies greater than 10 MHz. The greatest effect is achieved in power amplifiers with push-pull output stages.
Author: Gunderson, Steven J.

[lumanauw]

The Rotel's schematic looks like a cascoded VAS, but towards output node, not static cascode.
The patent's intention seems to isolate the output stage (distortion+nonlinearities) from affecting the input/previous stage by putting a common base transistor between output stage and VAS.
John Curl have the same intention, but doing it by putting a Mosfet predriver.

[mikeks]

Originally posted by lumanauw
The Rotel's schematic looks like a cascoded VAS, but towards output node, not static cascode.

Actually, Rotel's implementation is identical to Gunderson's-in his JAES paper.

[Zen Mod]

http://www.pat2pdf.org

just in case.....I bookmarked it long time ago....

[djk]

ITEM # PCBV-2 Lang 20W Mosfet Power Amplifier
From Audio Amateur Issue 2/86. Lang 20W single-channel Mosfet power amplifier. 5-11/16" x 6-1/16". 1 lb. $21.40

http://www.audioxpress.com/bksprods/pcbs/solidstate.htm

Uses lateral FET outputs, fully comp, in to out.

[mikeks]

Originally posted by djk
Lang 20W Mosfet Power Amplifier

Does it use the same compensation?

[janneman]

Originally posted by mikeks


Does it use the same compensation?

No. The Lang is classic: diff input bipolar, cascoded Vas, SK133/SJ48 output stage. Only comp is 68pf C-B on input side of diff input pair, probably because of the low ol gain (heavily degenerated input pair).

Jan Didden

[MikeBettinger]

Originally posted by mikeks

Actually, Rotel's implementation is identical to Gunderson's-in his JAES paper.

I'm curious as to when the JAES paper was published (and where I can get a copy of it to read ) I've been using this form of cascode for years after seeing it used in the GAS Ampzilla II (and IIa). The patent was granted in April of 1985, and I believe The Great American Sound company was dead and gone in 1981...

Sonically This is a great circuit. It is a lot purer in the highs and it also removes a bit of a resonant heavyness noticed most at the bottom end of the mail voice. To my ear it sounds right.

Mike.

[Zen Mod]

Originally posted by janneman



No. The Lang is classic: diff input bipolar, cascoded Vas, SK133/SJ48 output stage. Only comp is 68pf C-B on input side of diff input pair, probably because of the low ol gain (heavily degenerated input pair).

Jan Didden

hit us with schm ,just for fun

[darkfenriz]

Originally posted by mikeks
What do y'all think of Steven Gunderson's approach?

Note that this is the technique used by Rotel in their RB1090 power amp., and Randy Slone in his designs.

Also available as:

'A Topology to Linearize Miller-Effect Compensated Amplifiers'

JAES Volume 32 Number 6 pp. 430-434; June 1984
An additional topology to Miller-effect compensated amplifiers, which reduces high-frequency distortion, is described. The addition is implementable in discrete or integrated circuits, with little or no additional complexity required in designs incorporating short-circuit protection. This topology is effective in reducing distortion to frequencies greater than 10 MHz. The greatest effect is achieved in power amplifiers with push-pull output stages.
Author: Gunderson, Steven J.

In fig. 5 the emitter of Q24 'common base transistor' is at constant voltage against output, so Q24/Q25/Q26/diodes/CCSs circuit is a follower by its 'lovemaking' definition, it just consumes considerable DC comparing to classic followers.
Aside from this I see two poles and a local phase lead back to LTP, so a transfer function has 2 poles and one zero, not so uncommon I think.
Agreed? Please Mike don't answer just 'Nope...'
I would be interested on load stability of such a circuit.

best regards
Adam