[Chris Hornbeck]

Before this gets any sillier, let's all do a little
mental exercize. Let's draw two pictures of our
imaginary amplifier's response and view them together.

Our first picture is of the amplifier's "magnitude"
(just a way of saying frequency response) response,
and is drawn log-log, meaning amplitude (x) is in dB's
and frequency (y) is in decades or octaves.

Our second picture is of the amplifier's "phase"
(just a way of saying differentiated group delay),
response and is conventionally drawn with linear x
("degrees" of phase) and on the same log y scale as
our first picture.

Buckeroos and buckerettes, you've just drawn a Bode
diagram. Nothing more and nothing less. And you can
no longer say that it's some Cabalistic secret.
That's it, period. There's *nothing* else.

Now draw another one for the feedback path. Now draw
a third for the *combination* of the two, noting that
you can simply add values in the x axis to get the sum.

The resultant diagram is the relevant one when discussing
stability. Still small-signal, yada yada, as proposed
earlier, but will be equally applicable for large-signal
and extreme loading as discussed later.


Now, does anyone have anything useful to contribute WRT
a modern method of automating data collection? Computer
soundcards would appear the obvious choice, but we'll
need one that can be subverted into a higher frequency
response than their intended use.

All thoughtful input requested,

Chris Hornbeck

[Patrick Turner]

Chris Hornbeck wrote:

Before this gets any sillier, let's all do a little
mental exercize. Let's draw two pictures of our
imaginary amplifier's response and view them together.

Our first picture is of the amplifier's "magnitude"
(just a way of saying frequency response) response,
and is drawn log-log, meaning amplitude (x) is in dB's
and frequency (y) is in decades or octaves.

Our second picture is of the amplifier's "phase"
(just a way of saying differentiated group delay),
response and is conventionally drawn with linear x
("degrees" of phase) and on the same log y scale as
our first picture.

Buckeroos and buckerettes, you've just drawn a Bode
diagram. Nothing more and nothing less. And you can
no longer say that it's some Cabalistic secret.
That's it, period. There's *nothing* else.

Now draw another one for the feedback path. Now draw
a third for the *combination* of the two, noting that
you can simply add values in the x axis to get the sum.

The resultant diagram is the relevant one when discussing
stability. Still small-signal, yada yada, as proposed
earlier, but will be equally applicable for large-signal
and extreme loading as discussed later.

Now, does anyone have anything useful to contribute WRT
a modern method of automating data collection? Computer
soundcards would appear the obvious choice, but we'll
need one that can be subverted into a higher frequency
response than their intended use.

All thoughtful input requested,

Chris Hornbeck

Without being able to view a sample diagram, you've completely lost me
at about the third paragraph.

Buckerette? Hmm, country gal who does it?

What's her phone number?

Getting laid would be easier than understanding stability.

And I don't have to avoid tempting buckerettes, or even La suckerettas
from Mexico,
they sure avoid me.

Patrick Turner.

[Patrick Turner]

flipper wrote:

On Tue, 15 May 2007 11:09:45 GMT, Patrick Turner
wrote:



Chris Hornbeck wrote:

Before this gets any sillier, let's all do a little
mental exercize. Let's draw two pictures of our
imaginary amplifier's response and view them together.

Our first picture is of the amplifier's "magnitude"
(just a way of saying frequency response) response,
and is drawn log-log, meaning amplitude (x) is in dB's
and frequency (y) is in decades or octaves.

Our second picture is of the amplifier's "phase"
(just a way of saying differentiated group delay),
response and is conventionally drawn with linear x
("degrees" of phase) and on the same log y scale as
our first picture.

Buckeroos and buckerettes, you've just drawn a Bode
diagram. Nothing more and nothing less. And you can
no longer say that it's some Cabalistic secret.
That's it, period. There's *nothing* else.

Now draw another one for the feedback path. Now draw
a third for the *combination* of the two, noting that
you can simply add values in the x axis to get the sum.

The resultant diagram is the relevant one when discussing
stability. Still small-signal, yada yada, as proposed
earlier, but will be equally applicable for large-signal
and extreme loading as discussed later.

Now, does anyone have anything useful to contribute WRT
a modern method of automating data collection? Computer
soundcards would appear the obvious choice, but we'll
need one that can be subverted into a higher frequency
response than their intended use.

All thoughtful input requested,

Chris Hornbeck

Without being able to view a sample diagram, you've completely lost me
at about the third paragraph.

Here ya go

http://en.wikipedia.org/wiki/Bode_plot

Here's a page that let's you play with some. Woohoo!

http://circuitscan.homestead.com/fil...ircp/bode1.htm

Well thanks for the references.

I draw curves like these all the time when checking amplifiers, and
fiddle with
networks to improve stability, so probably I am intuitively making and
interpreting Bode plots
which i don't call bode plots. They are just my
recorded plots of gain and phase shift.

Patrick Turner.




Buckerette? Hmm, country gal who does it?

What's her phone number?

Getting laid would be easier than understanding stability.

And I don't have to avoid tempting buckerettes, or even La suckerettas
from Mexico,
they sure avoid me.

Patrick Turner.

[Iain Churches]

"Chris Hornbeck" wrote in message
...
Before this gets any sillier, let's all do a little
mental exercize. Let's draw two pictures of our
imaginary amplifier's response and view them together.

Our first picture is of the amplifier's "magnitude"
(just a way of saying frequency response) response,
and is drawn log-log, meaning amplitude (x) is in dB's
and frequency (y) is in decades or octaves.


OK. I can do that

Our second picture is of the amplifier's "phase"
(just a way of saying differentiated group delay),
response and is conventionally drawn with linear x
("degrees" of phase) and on the same log y scale as
our first picture.

OK. I can do that too

Buckeroos and buckerettes, you've just drawn a Bode
diagram. Nothing more and nothing less. And you can
no longer say that it's some Cabalistic secret.
That's it, period. There's *nothing* else.

Now draw another one for the feedback path. Now draw
a third for the *combination* of the two, noting that
you can simply add values in the x axis to get the sum.

Hmm. This is where it gets interesting.

Now, does anyone have anything useful to contribute WRT
a modern method of automating data collection? Computer
soundcards would appear the obvious choice, but we'll
need one that can be subverted into a higher frequency
response than their intended use.

Chris. I have bench instruments, rather than a soundcard for
measuring. I can measure and plot both level and phase.

Can you talk us through a worked example.
Excellent thread - thanks in advance.


Cordially,
Iain

[Ian Iveson]

Chris Hornbeck wrote more than that quoted below.

Glad "us all" can have a transfer function in the feedback path now,
unequivocally.

But this isn't clear...

Now draw another one for the feedback path. Now draw
a third for the *combination* of the two, noting that
you can simply add values in the x axis to get the sum.

Isn't the y axis supposed to be the vertical one? And what does "add
values in the x axis" mean anyway? Might help also to imagine a y1 and
a y2 axis...

I assume you mean add the heights of the log "magnitude" graphs at
each frequency to get the product of the "magnitudes", and the same
for the "phase". This is the kind of thing that makes sense if you
know it already, but is open to serious misinterpretation if you
don't, because adding logs is multiplying values. Transfer functions
multiply in series.

Also, it may be worth mentioning that this method of combining bode
plots does not depict the whole caboodle...only the part that is
relevant for stability.

Surplus dual CROs are cheaper than decent soundcards maybe, if you
don't have to pay carriage. Depends what context you have in mind,
particularly quite what you wish to automate. A computer is the only
general-purpose automaton most people have.

I wonder again who you are aiming this at?

All thoughtful input requested.

Thoughtful but not necessarily input.

Thanks, Ian

[Chris Hornbeck]

On Tue, 15 May 2007 14:40:55 GMT, Patrick Turner
wrote:

I draw curves like these all the time when checking amplifiers, and
fiddle with
networks to improve stability, so probably I am intuitively making and
interpreting Bode plots
which i don't call bode plots. They are just my
recorded plots of gain and phase shift.

Bingo! Oh ye of little faith...

Much thanks, as always,

Chris Hornbeck

[Chris Hornbeck]

On Tue, 15 May 2007 19:21:20 +0300, "Iain Churches"
wrote:

I have bench instruments, rather than a soundcard for
measuring. I can measure and plot both level and phase.

Me too. The soundcard automation my (likely) not
turn out to be possible anywho.

Can you talk us through a worked example.

Cool; let's do a separate thread after the (several)
bugs are worked out of this one.

Much thanks, as always,

Chris Hornbeck

[Chris Hornbeck]

On Tue, 15 May 2007 18:01:21 GMT, "Ian Iveson"
wrote:

But this isn't clear...

Now draw another one for the feedback path. Now draw
a third for the *combination* of the two, noting that
you can simply add values in the x axis to get the sum.

Isn't the y axis supposed to be the vertical one?

Yes; very sorry; I never actually label them; sorry for
the poor proofreading.

And what does "add
values in the x axis" mean anyway? Might help also to imagine a y1 and
a y2 axis...

I assume you mean add the heights of the log "magnitude" graphs at
each frequency to get the product of the "magnitudes", and the same
for the "phase". This is the kind of thing that makes sense if you
know it already, but is open to serious misinterpretation if you
don't, because adding logs is multiplying values. Transfer functions
multiply in series.

Logs are used for the *Y* axis because the curves become
(asymtotically) lines and because it's possible to simply
add inches to calculate values.

If one *Y* level is 1.4 inches below reference and the *Y*
level being added is 0.6 inches above reference, we need
only measure down 1.4 inches, then back up 0.6 inches (or
do the math first! arf!) and plot of result for that
frequency. It's ancient and graphic.


Another correction: although it should be obvious, feedback
must *first* be removed from the input summing point
before measurements. Suitable artificial loads are
substituted for circuit loading, as needed.

This wasn't spelled out in the original post, and
could have been misinterpreted.

Much thanks, as always,

Chris Hornbeck

[Ian Iveson]

I probably stupidly said

and the same for the "phase"

Er, no that wouldn't be right. Magnitudes multiply and angles add, so
the magnitude axis should be log and the phase axis linear. Then you
can add the graphs. Some of my errors only become apparent late when
I'm unwinding, and it's too late.

cheers, Ian

[Ian Iveson]

"Chris Hornbeck" wrote in message
...
On Tue, 15 May 2007 18:01:21 GMT, "Ian Iveson"
wrote:

But this isn't clear...

Now draw another one for the feedback path. Now draw
a third for the *combination* of the two, noting that
you can simply add values in the x axis to get the sum.

Isn't the y axis supposed to be the vertical one?

Yes; very sorry; I never actually label them; sorry for
the poor proofreading.

And what does "add
values in the x axis" mean anyway? Might help also to imagine a y1
and
a y2 axis...

I assume you mean add the heights of the log "magnitude" graphs at
each frequency to get the product of the "magnitudes", and the same
for the "phase". This is the kind of thing that makes sense if you
know it already, but is open to serious misinterpretation if you
don't, because adding logs is multiplying values. Transfer functions
multiply in series.

Logs are used for the *Y* axis because the curves become
(asymtotically) lines and because it's possible to simply
add inches to calculate values.

Yes, which is multiplying magnitude values. There is some
justification for the log scale also in the fact that our hearing is
supposed to be log too. Sometimes it's nice how the world fits
together.

If one *Y* level is 1.4 inches below reference and the *Y*
level being added is 0.6 inches above reference, we need
only measure down 1.4 inches, then back up 0.6 inches (or
do the math first! arf!) and plot of result for that
frequency. It's ancient and graphic.

Napier's not so ancient. Graphical methods alone can be fatal to
understanding.

Another correction: although it should be obvious, feedback
must *first* be removed from the input summing point
before measurements. Suitable artificial loads are
substituted for circuit loading, as needed.

Well, this raises an interesting issue. If the feedback truly went to
a passive summing point at the input, then you wouldn't need to
disconnect it because you could measure the forward response to the
summed signal. With the usual method of returning to the cathode, you
could measure the summed input between cathode and grid, and compare
that with the output perhaps.

I feel I should mention again the point that adding your two plots
together does not result in a plot which describes the amp as a whole.
It is what you want for stability considerations though, so you're not
wrong. It's just I imagine some folk may think that the amp actually
has that response.

cheers, Ian