Re: squeezing a field

On 28/10/2024 11:06 pm, piglet wrote:

Bill Sloman <bill.sloman@ieee.org> wrote:

On 28/10/2024 3:43 am, legg wrote:

On Sun, 27 Oct 2024 15:01:34 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:
>

On 27/10/2024 2:20 am, legg wrote:

On Fri, 25 Oct 2024 16:08:08 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:
>
<snip>

He was remarkably good, just not totally perfect.
>
His footnote reference to "squegging" in the 1959 class-D oscillator
paper is another minor drop-off. He can't be blamed for it, but a
super-hero might have done better.

>
What's wrong with 'squegging' ? It's a simple word that covers
a host of faults that all give the same approximate symptom . .

>
With the advantage of 65 years of hindsight, it looks as if what he was
seeing was gain in bipolar transistors running in the inverted mode.
>
"Squegging" was mostly used for  weird oscillations in resonant circuits.
>
Class-D oscillators built with MOSFet switches don't squeg. Class-D
oscillators built with bipolar transistors in LTSpice don't squeg either
- the Gummel-Poon transistor model doesn't model inverted mode behavior
all that well.

>
Squegging in any oscillatory circuit, driven or otherwise,
describes widely varying amplitudes that typically approach
self-quenching and can otherwise approach unintentional
overstess in the 'wobulating' cycle.
>
Not what the doctor ordered, or the designer anticipated.
>
Only blocking oscillators do it on purpose.

>
What Baxandall was describing was a situation where you've built a
class-D oscillator and used a feed inductor which has an appreciably
higher inductance than the inverter transformer.
>
If you simulate that in LTSpice, the voltage at the centre tap starts
off climbing up to about twice the steady-state peak and drops below the
rail during recovery, but this roller-coaster effect dies away. In real
life it doesn't (if you are using bipolar transistor for your switches).
>
My guess is that you could stop it by adding the right zener diode
between the centre tap and ground - one that didn't ever conduct when
the circuit was running smoothly, but would start conducting if the
centre tap got much above the steady state peak. This stops the
centre-tap ever getting below the rail at the bottom of the start-up
roller coaster - or at least it does in LTSpice and would keep you away
from the mode of operation where the switching transistors were
operating in the inverted mode.
>
Peter Baxandall invented the circuit before 1959, before Zener diodes
were widely available.

 That topology is very critical around conduction overlap vs dead band and
nano seconds matter.

If you drive the transistor bases with a centre tapped secondary (with many fewer turns) as Peter originally described, the nanoseconds look after themselves. Conduction overlap isn't a good idea but an handful of nanoseconds of underlap isn't a problem.
--
Bill Sloman Sydney