Hi Bob



 RM> Final cure was a teensy little .01uF across the collector-base

 RM> of the TIP driver.



 RM> I suspect the "real fault" was poor layout.



 BL>   It's never layout. These things take off around 1-10 MHz and layout

 BL> problems don't begin till the tens of Mhz. My guess is that a large

 BL> proportion of *all* those converters oscillated, depending on the

 BL> actual transistors fitted... a design fault.



It's a fault I haven't seen before, so I wouldn't call it a "large"

proportion



 BL>   The problem is that a transistor is a 2-pole circuit at HF: the base

 BL> resistance and collector-base junction capacitance, plus the emitter

 BL> resistance and the base-emitter junction... and to make it worse, the

 BL> usual trick is to drive one transistor straight out of the other, so

 BL> Miller effect adds yet another time constant. As soon as you have a

 BL> 3-pole R-C circuit you have a possible oscillator and none of the

 BL> elements are linear, varying with voltage and load current... and

 BL> load capacitance.



So any Darlington pair is a potential HF oscillator? This circuit

was quite simple, a 741 drove a TIP31C through a 470 ohm, the TIP

drove three parallelled 3055s, each of which had a small emitter

resistor. The source impedance didn't alter things (there were

about 10 volts of rf on the collectors at first, extra bypassing

reduced that to well under 1 volt, but the output changed very

little), the load impedance was low, the current limit circuit was

innocent (removing it altogether changed nothing), and the 741 had

nothing to do with it. I've seen amplifiers of all sorts take off,

and the problem is either poor design (read: layout) or some sort

of impedance where it shouldn't be, like a grotty battery terminal

in a portable radio making the audio amp take off.



 BL>   That 0.01 you slap on is a real bastard. It starts a top cut at

 BL> around 1KHz with the idea of reducing the gain to zero at 1Mhz where

 BL> the pole exists, but it also adds another 90-degrees phase shift that

 BL> can sometimes make it worse. So you add another 0.01... somewhere, and

 BL> then a spoiling 0.1uF/10 ohms across the output, and so on...



I'm quite aware of the risks of sticking a low freq pole partway

down a feedback amp. In this case, however, it worked. QED.



Cheers



--- PPoint 1.88