Ian Iveson wrote:

The OP asked in the context of valve amps. I do work on valve amps
exclusively, and they are usually made by
me, and they have grid stoppers where they should have.

If I use overall nfb, then ringing for a power amp is likely to be
around 80kHz and 1MHz. Never thought through why this always seems
the case. The former is dealt with by snubbing the OPT primary
somehow, and the latter by power-valve grid stoppers usually.

The reason for oscillations in this 80 kHz to 1MHz region is due to the
accumulated phase shift of the voltage amp, and that of the OPT,
which acts like a second order LPF, due to the series leakage
inductance,
and shunt capacitance.
As one rolls the F upwards, the open loop phase shift might be say
180 degrees at say 150 kHz.
Even though the gain has rolled off a lot, its sufficient that when FB
is applied,
the fed back signal is positive FB, and without an output load, gain is
at a
a maximum, and maybe it just oscillates. With a resistance load,
maybe it stops oscillating, but will still oscillate if a cap load of
typically
0.22uF is added. The 0.22 causes a further phase shift.
The use of just the right values of R&C connected across the load of V1,

and the OPT half primaries, will usually reduce the open loop gain at
HF,
but not below 20 kHz.
Also, the zobel network at V1 reduces the phase shift between V1 and V2,

whatever they may be, and although phase shift at 20 kHz is slightly
increased
in the open loop response, the stepped response of V1 due to the
load value being dropped from say 47k down to say 4.7k means the miller
effect between
V1 and V2 is reduced, and the full 90 degrees of ultimate phase shift
will be moved from around say 100 kHz to say 500 kHz, by which time the
V1 gain
has been reduced say 12 dB by the zobel.
This is critical damping for a bandpass filter with feedback, and all
amps are bandpass
filters, as well as amplifiers.
The cap across the Feedback Resistor to the cathode of V1 from the
OPT secondary, if this is fitted, is chosen with caution.
As the F rises, the phase of the fed back voltage applied to the V1
cathode
is advanced, to compensate for the lag in phase by the time one gets to
say 150 kHz,
The fedback voltage thus is more in phase with the input voltage,
so the feedback remains effective up to a higher F.
Too big a compo cap, and phase shift at say 500 kHz becomes positive,
and voila, the amp oscillates at around 500 kHz.

With SS amps, there is no OPT, and no leakage inductance,
but there is miller effect in the driver amp, and phase shift in the
whole amp, but usually the the first HF pole is at 1 kHz, and open loop
response falls at
6 dB/octave above 1kHz, so the open loop response is down 60 dB at 1MHz.

if 40 dB of NFB is applied, then the response is flat to say 100 kHz.
a zobel network across the main gain stage of the SS amp is still
required,
to prevent early phase shift, and lower the open loop gain,
so when FB is applied, you don't have an RF transmitter.
I am simplifying what happens here, and most SS amps have
inductive zobel networks operating between the output and the speaker
terminals,
and feedback loop does NOT include this passive network,
which prevents the amp from ever suffereing the phase shift due to being
directly
connected to a capacitive load.
Sometimes SS amps oscillate, but usually at much higher F than tube
amps.

All amps using NFB have to be set up carefully if they are to remain
stable under all load conditions, including connection of a 0.22 uF
across the output, with no R load.
This makes many tube amps oscillate violently.



If I built pre-amps with feedback then a better scope might attract
me more. As it is I rely on sensible precautions.

In the fullness of time, if you do want to see more of what happens with

amps that have NFB, or oscillate by means of stray capacitance between
an output and an input someplace, then a 20 MHz cro is
probably all you really need for most tube circuits.
Occasionally, one does see RF oscillations at higher F than 20 MHz,
and a sixth sense is needed to detect them, or a meter.
A cascode preamp I recently built did oscillate at some high RF,
and just touching parts where the voltage was supposed to be fully
bypassed
with a screw driver caused a loud click at the output speaker.
the screw driver changed the effective circuit at RF, which is quite
different
to what you have at audio, and the oscillation stopped, and the
stop and start of a stream of oscillation will cause a click, that of
course shouldn't
be heard.
Plenty of info about simple RF detectors in ARRL handbooks.

Patrick Turner.



cheers, Ian

"Phil Allison" wrote in message
u...

"Ian Iveson" wrote in message
...

"Phil Allison" wrote


** If you can't see them on the scope how do you even know
when
they are there ?

I measure with a voltmeter. Also, the scope's trace goes fuzzy.


** That contradicts my "if" condition .

10 Mhz makes no effect on a 1 Mhz scope.



If you don't know they are there then why would you use
an
(RF) voltmeter ??

To check they are not there :-)


** You are assuming continuous RF oscillations - more usually
ones sees
bursts superimposed on a audio wave.



This is in the context of audio valve amplifiers. When I check
an
amp I put a 25MHz voltmeter across the load. If I get a
significant
reading with no input signal, it is oscillating. If I need to
know
the frequency of oscillation then I use a frequency counter.
The
scope only gets connected when I know the amp is not
oscillating.
Both these instruments are decent, ex-military units and fairly
recently calibrated. I can read them without using a ruler.


To check for ringing above 2MHz, I suppose I could sweep and use
the
voltmeter, but I don't bother. Perhaps for guitar amps, or
other
things with complicated voltage stages and crowded wiring?


Do you often discover anything worth finding above 2MHz? Maybe
I
should save up for a new scope?


** I started out ( as a teenager) with a home brew 3 inch, 2
MHz all
tube scope 35 years ago ( still have it too) then a 10 MHz
single beam and
for the last 16 years a 50 MHz dual beam.

SS amps can have parasitic oscillations up to 40 MHz or more,
especially
if they use MOSFETS.

Probably 2 MHz is enough for someone who works on tube amps
exclusively.




............. Phil