Re: NFB 101 Part Deux



Ian Thompson-Bell wrote:

Iain Churches wrote:
"Ian Thompson-Bell" <ruffrecords@xxxxxxxxxxx> wrote in message
news:frh1t4$e34$1@xxxxxxxxxxxxxxxxxxxx
Sorry it has been a while coming but I have just finished part two of NFB
101. I found I could not describe it properly without diagrams and proper
equations so I had to resort to using a word processor and converting the
result into a pdf, all of which took time. So point your browser here:
http://www.ianbell.ukfsn.org/data/nfb101.pdf to download it.

Excellent Ian. I have just printed this out for bedtime reading.
And just as Andre had reported there was nothing happening.
Your timing was impeccable:-)

Iain



There may be typos in it still. I changed all the text based formulae to
'proper' ones so something may have got lost in translation. There are
certainly some superfluous brackets lying around only because I have not
yet worked out how to get the formula editor to suppress them. And I am
not sure if the figures tie up exactly with the text and of course it is
nowhere near finished yet.

Cheers

Ian

Good work Ian.


RDH4 has all this spelled out of course, and uses
µ , "mu" for amplification factor and ß "beta" for fraction of output
voltage
either in series with input voltage for series voltage NFB or in shunt
with applied NFB
as in shunt voltage NFB.

RDH4 lists all the many varieties of feedback, both positive and
negative,
and current and voltage types and whether it is shunt or series.
RDH4 also has a table to show what the effects are of all the different
types
of NFB.

But not every single fascinating aspect of feedback is explored in RHD4.

For example, did you know that positive "series" current feedback will
reduce the output
resistance of any amplifier, but at the cost of reducing bandwidth and
increasing distortions?

If we consider a "normal" amp response of having a reducing output
voltage as the RL becomes lower
as having a POSITIVE output resistance, then the positive CFB can reduce
this amount of Rout
to a lesser value; a typical ultralinear amp with say Ro = 7 ohms with
PCFB
can easily have its Rout reduced to say 1 ohm.
We assume Rout has been reduced from +7 ohms to +1 ohm.
Then the application of the global series voltage "normal" NFB will
reduce it even more to a lower value along with all distortions of the
amp and those
produced by the PCFB applied internally.

We can even apply so much PFFB that the positive Ro becomes a NEGATIVE
output resistance,
and the result is that output voltage rises when RL value is reduced.
So +1 ohm can be reduced further so Rout = ZERO ohms, then even less
than zero ohms, ie
maybe -1ohm.
Obviously, such negative Ro tempts fate and credulity because its
impossible
to get an amp which makes say 16Vrms at clipping into 8 ohms to make
20Vrms into 1 ohm.
But it is possible to make the same amp which makes 1.6V into 8 ohms
deliver 2V into 1 ohms.
But we would find this difficult to live with; once you examine how its
done, and experiment
with it, there are definate stability issues, and the open loop gain (
OLG ) must be reduced or phase tailored
to prevent oscilations, and the application of the PCFB be prevented at
extremities of LF and HF bandwidth.
Very interesting, but don't say I have not warned you.

The maths involved around each different form of feedback will take you
another 20 pages to explain.

I do suggest that all your terms for RL, Ra, and all others be made the
same as in RDH4,
because it was a good standard and everyone should know it, and that
equations be written in the same way.

In your theoretical workings for NFB application, how about showing some
typical
tube power amplifier schematics with NFB applied with all working
voltages with their polarities
so ppl can measure their own amps and understand it all a lot better?

Using a triangle pointing to the right to represent an amp with two
inputs on the
left vertical side and one output at the right point is the text book
way to represent an amp
so people do not have to keep in mind all the complex inner amp topology
which distracts them from the basic idea.
The same model can then be used for a tube amp or an opamp.

However, to include all possible phase shift peculiarities of the open
loop
character of an amp and the equivalent networks in the amp which produce
them
and their interaction when FB is used takes rather a lot of work.

NFB theory and application has already covered in many old books, and
many should be found then read,
and the messages in each will overlap each other books's shortcomings,
and you end up wize while you
remember it all, then dumb again when you forget it all.
Unless of course all you do all day everyday is design and stabilise new
and old amplifiers.
I probably do enough to keep me wize.

I have never seen an online calculator for NFB, where one dials in the
details of the open loop gain and all its phase shift rates and bothers,
and then
ask for 20dB of NFB, and click "calculate", and have the program come up
with the FB network and including all the phase tweaking networks needed
for
unconditional stability into any possible type of reactive loading, R
load, or no load at all.
Such a program could possibly be a boon for the dumbos to whom feedback
is a terrible mystery, and always will be,
and hence hated fiercely, and avoided.
But with a programmed or synthesized solution,
one must ensure it is still a viable solution which works in practice.
Since garbage in = garbage out with simulation programs, expect many
simulated solutions using tubes and OPTs to still be good oscillators
when nobody expected it.
That's because its difficult for anyone to correctly define all the open
loop
gain and phase shift character. Too fucking hard.
Just bulid it, and learn to stabilise it by empirical methods of network
applications and trial and error
and by observation with a CRO. This is a far quicker way than all the
calculations in the world sitting
down at a table when you should be in the workshop achieving something
real.

Models of the single tube amplifier stage should include an extremely
low voltage generator producing output of µ x Vg
with series resistance between the gene output and the anode terminal
should be explained
as equivalent models of the triode or pentode ot any other tube.
Newbies NEED to know the very boring basics before thay can have any
chance of understanding.
Most don't have a clue what a voltage generator is, or what the dynamic
anode resistance is at all!!!

I welcome you to borrow whatever you need from

http://www.turneraudio.com.au/tube-operation1.html

and

http://www.turneraudio.com.au/tube-operation3.html

I don't have all the possible various feedback applications mentiuoned
at my site.

It is to be expected that the learning tube crafter will absorb the
basic ideas and
be able to perform all the basic calculations for normal series and
shunt VOLTAGE NFB.

The equations for gain with FB applied are slightly different for the
these two main types of NFB.

Since NOBODY ever dares to apply any positive voltage or current
feedback anywhere anymore to any amp whatsoever,
I have not mentioned exactly how it can be done successfully, because
the risks of
parasitic oscillations are so likely in the hands of ignorant would be
experts.

In general with low µ triode input/drivers, tube amplifiers have only a
mild amount of voltage gain without
any NFB applied, ie, "open loop" gain.
And this "OLG", open loop gain is then reduced by the OPT voltage ratio.

So, to force the amp to have an apparently larger amount of open loop
gain,
some positive voltage FB or positive current FB was applied and the
resulting amp
with mild applications of standard series voltage global negative FB
measured far better than with
just standard global series voltage NFB applied alone.
Its all mentioned in RDH4.

There was classic application of combined forms of FB in a commercial
example in Bogan amps
which had variable positive FB to vary the damping factor of their amps
and it gave better
sounding bass we have been told.

But such fiddles with FB or active manipulations of Rout are to be
treated with utmost caution.

With too much positive current FB, a shorted output can make an amp
oscillate
internally very violently, and cause its demise from thermal runaway
of the output tubes, so too much PCFB is extremely undesirable.
A little dab of PCFB goes a long way and is extremely effective in
reducing Rout a lot though
without increasing the bad artifacts of reduced BW and distortions very
much.

In SS amps, nobody ever would use old fashioned tricks of positive FB
anywhere
because the total including the input LTP, VAS and output BJTs of
mosfets is over 100,000x,
or more than 80dB, so when a TOTAL amount of series voltage NFB is
applied
in the form of say 30dB emitter or source FB in the output stage and
50dB of global NFB,
the gross distortions of crossover and THD and IMD in the OLG
charateristic are reduced
to well below what is ever possible to be heard.

Well, in theory that is, because often some ratbag pissant 20 watt SE
tube amp will
give a clearer rendition of what Motzart or Led Zeppilin intended than
some
100W SS thinge, and despite the fact the SE amp will have very little
NFB and
have 100 times the measured THD/IMD than the SS amp.

Numbers ain't everything.

Patrick Turner.
.