Four for SuSy?

[Onvinyl]

Is this Idea a valid option worth trying?
Improved linearity through complementary differential further tamed with sort of double-x?
Or would the circuit blow and explode due to the different charges meeting at the gates?



Just a silly morning-idea before coffee & work?
yours,
Rüdiger

[Nelson Pass]

Perfectly fine idea.

[GRollins]

Have a look at some of John Curl's designs as far back as the '70s and you'll see the same thing. Others seem to know it as a 'diamond differential' and I believe there's yet another name for it, but I can't remember what it is at the moment. I tend to think of it as John's, since that's where I saw it first, but I don't actually know who came first.
Yes, it will work for X circuits, although existing JFETs don't have a very great voltage range, so be prepared to cascode if you want more than a few volts of output.

Grey

EDIT: Caps in the feedback loops will take care of DC problems.

[Nelson Pass]

Actually, it looks very much like the X (not XA) front end. I believe,
though, that the diamond differential front end was a design
by Sansui to get very high slew rates. I have not seen it in
a product since, but perhaps people are applying the name to
what I would refer to as Curl's JFET dual differential. It seems
to me that dual differentials had already been invented, but
John used the Vgs character of the JFETs to self-bias very
elegantly, much as you might see in tubes, but symmetric with
P and N channel devices.

[GRollins]

John posted here on the history/taxonomy of the circuit some months back, but I'm not in the mood to argue with the search engine to try to find it. For some reason they frown on alcohol here at work and I refuse to wrestle with the search function without something liquid to sustain me.
From memory: I think he said he came up with the idea at Ampex(?) in the late '60s. Someone else came up with the same general thing at about the same time. The Sansui version was in the mid-to-late '70s; they drew it differently, but it boiled down to the same thing. If someone knows where John's post is, it would be better to rely on his rendition rather than my recollection of his post.
At any rate, you can go to www.marklev.com and look at the schematics of the JC-2, where he used it in the line stage.
I've played with the circuit with excellent results and I'm sure that Nelson has flogged it within an inch of its capabilities for the X amps. I believe Nelson cascodes his JFETs with bipolars, presumably Zetex parts.

Grey

[Nelson Pass]

I do indeed.

But with regard to the Diamond Differential, I don't recall the
exact schematic, but it is designed to kick in with lots of
additonal current beyond the bias point. It never achieved
popularity, I beleive because its sole purpose was to beat
the **** out a single spec - slew rate, which I regard as a
red herring.

(insert slew rate rant here)

JC's differential circuit was a model of subtlety and simplicity
by contrast, and lives on as a classic.



BTW, I did generate a bunch of performance curves on a circuit
like the one that started this thread. I'll dig them up.

[mlloyd1]

finally, a question for which I remember the answer

the link grey refers to is here:
double vs diferential

mlloyd1
who is looking forward to seeing nelson's curves
hmmm, that didn't come out right did it?

Originally posted by GRollins
John posted here on the history/taxonomy of the circuit some months back, but I'm not in the mood to argue with the search engine to try to find it. ...
Grey

[Onvinyl]

Hi,
Would be great to have that performance curves!
My question arouse when i was looking at the borbely circuits and was thinking of a way to make them 'real balanced' (apart from simply doubling them). Since those are (often) feedbacked via the '-' input node I recalled the 'x'-thing and re-read the zen/susy paper instead of a late night crime-fiction
If it all fits, I might use it as a second stage in a phono amp (after a passive network), so apart from very good linearity it is interesting how big the input voltage range is and how immune to complex input 'loads' (I don't know how to say that better).

But apart from my intention I'm eager to follow your input,
thanks,
Rüdiger

[Nelson Pass]

Here we go. This is the same circuit, with 8 mA total
bias, 2.2 Kohm Drain resistors, and +/- 30 volt rails.

The open loop gain is 50, the CMRR is -100 dB, and two
curves are shown, but I forget what condition change there
was - my notes make vague reference to 6 mA bias also,
and that might be the difference, but more likely is a different
load resistance.

[GRollins]

Hmmm...the post mlloyd1 linked to is half of the stuff I remember, but not all. I seem to recall John discussing the 'diamond' circuit as well. Must have been a different post in a different thread.
I don't have a schematic of the diamond circuit at hand, so I'll take Nelson's word for it that it had extra current add-ons. He's in a better position to judge these things than I am, as he seems to have schematics for everything ever made.
Nelson, are those curves for the 389/109? If so, isn't 30V a little hot for the 109, or is the part more durable than the spec sheets would have you believe? If I recall, the 109 is a 25V part, which is wierd considering that the 389 is 40 or 50V according to the sheets I've got here (buried somewhere under all the other stuff...).

Grey