Booster circuit distortion

[tmadness]

I have been trying to design a opamp voltage booster circuit. I have settled on a CB voltage amplification stage and a darlington EF stage. The output circuit is single ended. I am observing distortion (or maybe this is instability?) in the amp on the rising edge of a waveform.
I don't know why this is happening. my guess is the feed forward cap C3 on the CB stage.


Design requirements:
Vout as close to 100v as possible.
Works across a wide range of loads, delivering upto 500 mA.
Solve for a DC case first, optimize circuit to work as fast a few KHz.

EDIT:
After getting breakfast in me I saw the issue: the CB stage is not biased properly.
Is there anything fishy about this circuit?

[T3sl4co1l]

What's v_out pulling down to (R5?)?  Just GND?  Ah yep, so the bipolar input is just out of range.

The diodes don't do anything, but if you want sinking, you'll need bias diodes and PNPs there.  Shouldn't need R1, R9 can do it all, or better yet a CCS pulling up -- also increases the open loop voltage gain of Q2 so a shunting R+C on the cb_out node may prove helpful.

Oh, C3 is wrong -- that's positive feedback; though not enough to cause instability, unless unlucky reactances cause it to oscillate (given the right lead lengths / node capacitances, and enough fT in Q2).  Better for it to go to GND instead of cb_e, and better to put a resistor in series (thus, the shunting R+C mentioned above).  At least, I suspect it was intended to be negative feedback (like C1); but Miller works between B and C, not C and E.

Tim

[tmadness]

That cap c3 gives me the heebie jeebies too. but without it the circuit oscillates.
I was told back in uni by my professor that a miller cap goes in between input and output of an amp so I put one on the E (in) - C (out) .
So the RC ought to go between gnd and collector.

[T3sl4co1l]

Your prof left out a critical part, that it's an inverting amp.  In fact since it's noninverting, it's the anti-Miller effect!  Aunty Miller?...

I think the effect is, it's somewhat bootstrapped at the emitter so doesn't do too much there (it's not properly bootstrapped like Cbe is in an emitter follower, Vo/Vin ~ 1), and at the collector it's still doing the same bulk loading effect.  It should be better to GND, where it's just doing an ordinary transconductance amp + load impedance.  Which is a very fair way of looking at this circuit, it's got a voltage into a current input, and a current output, transformed back to voltage by the node impedance.

Alternately, you could do some Miller effect by adding series base resistance, but that would just make things a little weirder, I think; the resistor allows a fast feed-forward path, while Ccb pushes it straight into the output node but without the benefit of negative feedback that works when it's common emitter.

Which I've kinda sorta done before, to tune the gain/impedance of an amp stage.  This is an example:



Despite being emitter input, the first stage has a modest impedance (apparently, high enough that a shunt 51R (in part) was needed for termination), and the C-B feedback divider sets the voltage gain.  (The two top loads act like a gyrator, with an emitter-follower output, further increasing GBW and reducing distortion.  The downside is all the large coupling caps to maintain bandwidth down to modest frequencies.

And, not like you'd expect, or want or need, to use peaking caps and stuff for a much more modest bandwidth booster circuit.  Simple is good!

Tim