LINEUP Input-VAS, idea

[lineup]

It is good to make use of input stage gain.
Because it is good linear gain.
This is an idea I have tested and had since years ago.
I call it new, because I have never came across this before

See diagram.
----------
Input may as well be 4 BJT and not necessarily symmetrical.

* First does not take gain from U5/U12 in input pairs.

* Second feeds U7/U14 to emitter of VAS transistors
This gives some extra gain. Curl has used it.

* Third is my idea. Uses a complmentary pair. For BJT VAS we need voltage sources to set current in VAS.
- This gives almost twice the gain of first circuit.
- Also voltage supply can be separate from input stage.
This can be of benefit.

[lineup]

JFET VAS is even more elegant.
We can remove voltage sources
and drive VAS directly from input collector resistors!

See diagram.

There are plenty more variants to employ this idea.
There are some minor drawbacks, too.
Have a try

[kenpeter]

I don't know about sane people, but this seems like a great idea to me.

[jam]

Lineup,

..............looks like a cool idea for a maybe a phono stage. You must have been smoking the magic weed again.

[Bonsai]

I use #2 with bipolars and it works well. I first saw this in a James Bongiorno amp.

[AKSA]

I've tried this topology, though with single diff pair input stage.

It has great speed, but thermal drifts make balance, and particularly offset control, very tetchy.

The degen resistor between the emitters of the Rush cascode is critical as it sets transconductance.

I had more success using the RC in the input stage, rather than as VAS.

Hope this helps,

Hugh

[Bonsai]

Are you refering to #2 Hugh? If so, did you have these problems in open loop of closed loop condition?

[forr]

Hi,
This VAS looks like a series differential pair with some degenaration, and the circuit at right of figure in post #1 is similar to a Rush configuration, as said Hugh, where a resistor between the emitters has been inserted.

[AKSA]

Hi Bonsai,

Neither strictly.... a single LTP, terminated with dissimilar resistive loads, driving a RC with degeneration. Sadly, I can't recall all the details, it was more than ten years back, but I do recall that in closed loop the sound quality was ordinary, even properly compensated, and the offset control was poor. I moved a few years later, at the suggestion of Keantoken in fact, to the use of the Rush Cascode in the input stage. This proved better, both in terms of stability and SQ.

I think it's possible to get almost any topology to work well, but there is no doubting that some are better than others. Making that call is tough; when do you stop development and move on?

[unclejed613]

looks like your input gm would not be very stable, since your LTP tail currents would be changing with the input signal.... LTP tail current should be stable, which is why CCSs are used. there are also stability problems when driving VASs in series like that. a stable operating point would be difficult to pin down here. it probably would work "ok" with real components because tolerance variations would unbalance it enough to stabilize, but using matched components will make it unstable. on the other hand allowing it to work in an unbalanced and more stable condition will also increase distortion a bit. so it's a tradeoff, distortion or instability.