Discussing audio amplifier design -- BJT, discrete

[Jon Kirwan]

On Tue, 2 Feb 2010 22:30:31 +0530, "pimpom"
<pimpom@invalid.invalid> wrote:

Jon Kirwan wrote:
On Sat, 30 Jan 2010 01:11:28 +0530, "pimpom" wrote:

snip
I like the biasing scheme mentioned by Jon and use it for all
my
designs except the early ones using germanium transistors,
though
I don't know the name either. The biasing transistor can be
mounted on the output transistors' heatsink for temperature
tracking.
snip

Okay. I'm giving this a little more thought -- as it applies
to temperature variation. The basic idea is that the two
bases of the two output BJTs (or output BJT structures) must
be separated a little bit in order to ensure both quadrants
are in forward conduction. With a "Vbe multiplier" in place
and with its own BJT tacked onto the same heat sink, the idea
is that the the Vbe multiplier's own voltage separation will
shrink as temperature rises, exactly in some proportion
needed to maintain the designed forward conduction
relationship of the output BJTs.

To be honest, this designed forward conduction mode may not
be critital. It might move a class-AB around a little within
its AB operation, for example, if the voltage tracking with
temperature weren't flawlessly applied. And that may be
harmless. I don't know. On the other hand, if tweaked for
class-A I can imagine that it might move the operation into
class-AB; if tweaked for lower-dissipation class-AB it might
move the operation into class-B; and if class-B were desired
it could move it into class-C with associated distortion.

There are several parts of the basic Shockley equation. One
is the always-in-mind part that includes a kT/q part in it
and relates that to Vbe. The other is the Is part and Eg is
the key there. So one thing that crosses my mind is in
selecting the BJT for the "rubber diode" thingy. Unless it's
Vbe (at 27C and designed constant current) and its Eg are the
same, even though it is a small signal device, doesn't that
mean that the variations over temperature will be two lines
that cross over only at one temperature point? In other
words, basically matches nowhere except at one temperature?

It seems crude.


You lost me for a while with the Eg term. You mean the emitter
transconductance?

No. Eg is the effective energy gap, specified in electron
volts. Eg (and Tnom, which is the nominal temperature at
which the Is used in the Shockley equation is given at) are
used to account for and calibrate the variation of Is over
the BJT's temperature. In other words, Is is a function of
T, namely Is(T), and not a constant at all.

If you solve the Shockley equation for Vbe and then look at
the derivative (partial, since Is is momentarily taken as a
constant) of it with respect to temperature, you will see
that it varies in the _wrong_ direction... the sign is wrong:

Id(T) = Is(T) * ( e^( q*Vd / (k*T) ) - 1 )

which becomes:

Vd(T) = (k*T/q) * ln( 1 + Ic/Is(T) )

The derivative is then trivially:

d Vd(T) = (k/q) * ln( 1 + Ic/Is(T) ) dT

which is a positive trend, very nearly +2mV/K for modest
Ic... but __positive__.

Does that make sense? It just is wrong. BJTs don't _do_
that. The figure is more like -2mV/K. So why is the sign
wrong?

Because that isn't the whole picture. "Is" also varies with
temp. As in:

Is(T) = Is(Tn) * (T/Tn)^3 * e^( -(q*Eg/k) * (1/T-1/Tn) )

where "Tn" is the nominal calibration temperature point.

The new derivative is a bit large. To get it onto a silly
post page with some chance that it won't sprawl for lines and
lines, I have to set up these math phrases.

Assume:
X = T^3 * Isat * e^(q*Eg/(k*Tnom))
Y = Tnom^3 * Ic * e^(q*Eg/(k*T))

Then the derivative is (if you use fixed-spaced ASCII):

X+Y
k*Tn*T*((X+Y)*ln( -------- )-3*Y) - q*Eg*(X*T+Y*T+Y*Tn)
Isat*T^3
-------------------------------------------------------
q * Tnom * T * (X+Y)

What a mess, even then. Here again, Tn is the nominal
temperature (in Kelvin, of course) at which the device data
is taken and Eg is the effective energy gap in electron
volts for the semiconductor material. Of course, 'k' is the
usual Boltzmann's constant, q the usual electron charge
value, and T is the temperature of interest.

Eg often defaults to around 1.11eV in spice, I think. For an
Ic=10uA and a stock Isat of about 1E-15, the figure comes out
to about -2.07mV/K in the vicinity of 20 Celsius ambient.

Which is the more usual value.

The "Is" term is the y-axis intercept, which isn't actually
measured, by the way, but instead extrapolated from measured
values elsewhere.

All this is the reason I was asking about the voltage bias
mechanism (that rubber diode/Vbe multiplier thing) and
selecting its BJT vs those in the output stage. (Which, if
PNP _and_ NPN are used, probably themselves do not vary the
same as either other, even, so there is another problem there
as well.) It fries my brain thinking about selecting
"perfect" parts for this.

Another issue I'm starting to wonder about is sweeping out
charges in the BJTs at higher frequencies and providing
sufficient drive current to do it quickly enough. But one
thing at a time, I guess.

Perhaps a short diversion into my own background may be
appropriate here. Shortage of funds and scarcity of good books
even for those who could afford them in a technologically
primitive environment kept me from delving deeply into
semiconductor physics when I started teaching myself electronics
over 40 years ago. I had advanced Math in college, but lack of
practice has made me very rusty. You're probably much better at
that.

Your own experiences sound very much like mine, except that
you _did_ something with yourself in this area when I did
not. Something I very much respect in you and disrespect in
me. I grew up poor enough that I had to literally live in
homes without walls and work the fields as a laborer child
(before laws today now prevent that, sadly in some ways good
in others) in order to eat and survive. So I understand
"shortage of funds" in my very gut. Perhaps what differed a
little is that I also was living near Portland and there was
a library system I could access, riding my bicycle as a kid.
And I would sometimes even take a bus and use the university
library (particularly the 5th floor where the science
subjects were located.) I scored an 800 on the math section
of my SAT and was rewarded with entry into a university
scholars program at PSU. However, I had to work to pay for
the classes and books and in the end I simply couldn't handle
all of it on my own. Without a dad (he died when I was 7)
and no family to help out, I couldn't manage to do everything
and get by at school, too. So I dropped out well before the
1st year completed. Everything I know is self-taught. It's
a commitment.

I have been honored by being asked by Portland State
University to temporarily teach as an adjunct professor,
though. And I did that for a few years until they could find
their replacement professors. I enjoyed it and I think I did
well. When I visited the department, last year after some
dozen years of absence, I was greeted in the hallways by many
others who I sometimes barely remembered with sincere smiles
and talks of those days. So I must have made some kind of
impression there.

Maybe a difference is that I've made the study of mathematics
a centerpiece for me. Besides, it's central to the work I do
so I can't really ignore it. But since I love studying it, I
would do it, anyway. None of this means I'm properly trained
in it. However, even these days I get to spend time almost
every month or two with sit-down time with an active
physicist to get some additional education in Lie Groups or
catastrophe theory or reflection spaces and manifolds, and so
on. I find I really love both finite and infinite group
theory work.

Over the years, I developed my own shortcuts and approximations
using mostly basic algebra, trigonometry and bits of calculus
here and there, blended with empirical formulas.

And here, most likely, is our fundamental difference. I
cannot remember things without understanding their deep
details. I lump this to my "autism." (I have two disabled
children on the spectrum, the youngest is almost exactly like
I was at his age.) When I took calculus at college, it was
all a blur trying to remember what was called what and how
they applied. However, if I _understood_ it deeply, could
picture it well, I could re-derive almost anything on the
spot when I needed it for a test. In other words, while most
of the other students appeared to simply take notes and keep
track of details (and shortcuts) many of which they'd
remember, I couldn't work like that. My memory was _zero_
for names of people, and similarly for names of math
formulas. I had to understand them viscerally and "see" them
well, in order to be able to remember the concept. However,
I still couldn't rememeber the specific "formula." Just the
concept -- the visualization, the image. That wouldn't
provide me with an answer to a problem, merely an approach
that "seemed right." So I would simply use that image to
guide me in re-deriving the formula from scratch, every time.
The upshot was that I took longer than most in completing my
tests, because I spent so much additional time quickly
running through the derivations of the rules I needed, but
where I answered the problems I got them right.

I've never been satisfied, as a result of my own limitations
here, to memorize shortcuts and approximations. It doesn't
give me "sight." They are useless to me, since I cannot use
them for any other derivations, since they are themselves
only blunted tools for specific purposes that cannot be used
to extrapolate anywhere else. Which then forces me to depend
upon a memory I don't have. What I need is to _see_ the
physics itself so that I can then derive those approximations
and shortcuts on the fly, deduced to the specific situation
I'm facing at the time.

In any case, the Shockley equation seems to hold fairly well in
practice for the purpose of bias regulation within the
temperature range normally encountered.

No, it doesn't. Because the SIGN is wrong!! The Vbe doesn't
rise with rising temperature, it falls.

Temperature tracking with
simple circuits like diodes in series or a Vbe multiplier cannot
be more than approximate.

That seems to be what I'm getting. One "lucky" circumstance
seems to be that the Vbe multiplier is supposed to produce
about two Vbes with k=2 in k*Vbe, just when there are two
Vbe's in the output structure. That way, if the Vbe in the
multiplier moves around with temperature, the multiplier
doubles it in just the right way to handle the actual two
Vbes in the output pair. If it had been needed to set k=3 or
k=1.5 or k= anything else, there'd have been a problem again
because they wouldn't vary together.

But this brings up the other problem I am talking about. If
Eg isn't the same figure, the slope over temperature for the
Vbe multiplier and the output BJTs won't be the same slope.
That means they can intersect at some temperature, but never
really be right anywhere else at all.

Worse than this is the fact that PNPs are used on one side
and NPNs on the other. They _cannot_ possibly vary their
Vbes in matched ways. It's got to be a nasty problem. And
it seems to argue, in my mind, for some modified version of a
quasicomplementary structure on the output. What argues
against it so much is that, again, the driving structure
before the quasi structure is driving two kinds of quadrants
and this means the cross-over area _must_ be nasty looking,
indeed.

Because of that, I searched around and found out that there
is a correction structure to fix the quasi crossover problem.
It appears to use something called a Baxandall diode, though
for now I haven't learned the details of how it does what it
does.

Such a device can sense only the
heatsink temperature

But which quadrant do you decide to attach it closer to?

and,.except under long-term static
conditions, that temp will almost always be different from Tj of
the output devices. That Tj is what needs to be tracked and when
the output transistors are pumping out audio power, that
difference can be tens of degrees.

I can believe it!

I've seen this as a modification. In ASCII form:

A
|
,---+---,
| |
| \
| / R3
\ \
/ R2 /
\ |
/ +--- C
| |
| |
| |/c Q1
+-----|
| |>e
\ |
/ R1 |
\ |
/ |
| |
'---+---'
|
B

We've already decided that R1 might be both a simple resistor
plus a variable pot to allow adjustment. The usual case I
see on the web does NOT include R3, though. However, I've
seen a few examples where R3 (small-valued) exists and one of
the two output BJTs' base is connected at C and not at A.

The above circuit is a somewhat different version of the Vbe
multiplier/rubber diode thing. The difference being R3,
which I'm still grappling with.

I've seen R3 used in that position too, but never gave it much
thought until you brought it up. Offhand I still can't see a
reason for it either. Maybe for stability against a local
oscillation? Perhaps taking some time to think about it will
bring some revelation. Or someone else can save us the trouble
and enlighten us.

It is often a small value, 10s of Ohms. It might just be
what you are talking about because I've often seen people
talk about needing a 33 ohm base resistor on emitter follower
BJTs to snub high frequency oscillations. So you might be on
the right track there. Hopefully, someone else does know and
will feel like saying.

But does anyone know, before I go writing equations all over
the place, why R3 is added? Or is R3 just some book author's
wild ass guess?

A possibility. But I wouldn't go out on a limb and call it that

Hehe.

Jon

This is all pressing me into studying the output structure
more, I guess. It basically looks simple when I wave my
hands over it, but I suspect the intimate details need to be
exposed to view. On to that part, I suppose.

Jon

 

[Jon Kirwan]

On Tue, 02 Feb 2010 13:01:18 -0800, Jon Kirwan
<jonk@infinitefactors.org> wrote:

Assume:
X = T^3 * Isat * e^(q*Eg/(k*Tnom))
Y = Tnom^3 * Ic * e^(q*Eg/(k*T))

Sorry, should be consistent in terms with:
X = T^3 * Isat * e^(q*Eg/(k*Tn))
Y = Tn^3 * Ic * e^(q*Eg/(k*T))

Jon

 

[Jon Kirwan]

On Tue, 02 Feb 2010 13:01:18 -0800, I wrote:

Because of that, I searched around and found out that there
is a correction structure to fix the quasi crossover problem.
It appears to use something called a Baxandall diode, though
for now I haven't learned the details of how it does what it
does.

Here's some articles I found on the web by a single author on
audio amplifier design:

http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205207238&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205601405&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205801115&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205917273&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=206103226&printable=true

The last one of the above links __mentions__ the Baxandall
diode.

In looking at those, there is this one also listed at the
bottom of the last article above. I haven't read this one
yet, but include it just the same:

http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205202120&printable=true

I need to read all of these, I suppose.

Jon

 

[Paul E. Schoen]

"Jon Kirwan" <jonk@infinitefactors.org> wrote in message
news:eg7hm51jl8ogq5k6f6v1k00mvmhmuqrej3@4ax.com...

On Tue, 02 Feb 2010 13:01:18 -0800, I wrote:

Because of that, I searched around and found out that there
is a correction structure to fix the quasi crossover problem.
It appears to use something called a Baxandall diode, though
for now I haven't learned the details of how it does what it
does.

Here's some articles I found on the web by a single author on
audio amplifier design:

http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205207238&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205601405&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205801115&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205917273&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=206103226&printable=true

The last one of the above links __mentions__ the Baxandall
diode.

In looking at those, there is this one also listed at the
bottom of the last article above. I haven't read this one
yet, but include it just the same:

http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205202120&printable=true

I need to read all of these, I suppose.

This stuff is a bit too intense for my taste, but a Dogpile search of
Baxandall Diode turned up this manual for an amplifier that uses tubes and
transistors, and has a Baxandall diode in the output stage. Its stated
purpose was to "improve symmetry" and appears to add an additional diode
drop for the PNP-NPN pair to match the NPN-NPN darlington on the top side.

http://www.wimdehaan.nl/downloads/technicalmanualnishiki41.pdf

This article states that a Baxandall diode made little change in linearity:
http://www.embedded.com/design/206801065?printable=true

There is also a Baxandall Tone Control circuit which is discussed here:
http://digitalcommons.calpoly.edu/eesp/14/

For my own purposes, distortion of anything less than about 1% is probably
not worth paying for or striving to achieve. For audio, my ears are not all
that good and I would probably welcome distortion in the form of non-linear
frequency response to compensate for degraded sensitivity at the high end.

And there is also the argument that any sound that is naturally produced
will have some significant distortion that is actually part of the
listener's experience. Whatever the acoustics in a given auditorium may be,
they contribute to the waveshape as it is received by the listener's ears,
and it varies depending on where one is seated. Sometimes added distortion,
such as an echo, may enhance the enjoyment of the music, and coloration due
to an imperfect amplifier might just as easily be perceived as pleasant
rather than objectionable. It is in fact distortion that causes an
audiophool to prefer the "warm" sound of a tube amplifier over a laboratory
grade solid state amplifier.

I am more impressed with amplifiers that are extremely efficient, such as
PWM amps. And for some types of test equipment that I have designed, I had
to deal with maintaining phase shift to better than 1 degree into a range
of inductive, resistive, or capacitive loads, and with outputs of power
line frequencies of 45-450 Hz, for voltage sources up to 300 VAC, and
current sources up to 100 amperes, at 50 VA to 300 VA or higher. And they
had to be able to withstand overloads and short circuits.

Paul

 

[Jon Kirwan]

On Tue, 2 Feb 2010 21:04:53 -0500, "Paul E. Schoen"
<paul@peschoen.com> wrote:

"Jon Kirwan" <jonk@infinitefactors.org> wrote in message
news:eg7hm51jl8ogq5k6f6v1k00mvmhmuqrej3@4ax.com...
On Tue, 02 Feb 2010 13:01:18 -0800, I wrote:

Because of that, I searched around and found out that there
is a correction structure to fix the quasi crossover problem.
It appears to use something called a Baxandall diode, though
for now I haven't learned the details of how it does what it
does.

Here's some articles I found on the web by a single author on
audio amplifier design:

http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205207238&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205601405&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205801115&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205917273&printable=true
http://www.planetanalog.com/article/printableArticle.jhtml?articleID=206103226&printable=true

The last one of the above links __mentions__ the Baxandall
diode.

In looking at those, there is this one also listed at the
bottom of the last article above. I haven't read this one
yet, but include it just the same:

http://www.planetanalog.com/article/printableArticle.jhtml?articleID=205202120&printable=true

I need to read all of these, I suppose.

This stuff is a bit too intense for my taste,

It's gravy to me. If only I had the good sense to be able to
tell if it is being comprehensively and accurately stated.

but a Dogpile search of
Baxandall Diode turned up this manual for an amplifier that uses tubes and
transistors, and has a Baxandall diode in the output stage. Its stated
purpose was to "improve symmetry" and appears to add an additional diode
drop for the PNP-NPN pair to match the NPN-NPN darlington on the top side.

http://www.wimdehaan.nl/downloads/technicalmanualnishiki41.pdf

I think that very same purpose is what I took it to mean,
too. In a quasi-complementary output stage, the gain curve
(less than 1 everywhere) shown over output would have to show
an interesting tweak, midrange. In class-B especially, from
one side it would look one way, from the other, somewhat
different. Simply because the two quadrants just aren't the
same structure. One uses two NPNs, the other an NPN and a
PNP. That weirdness in gain has to translate to distortion
of some kind. The fix, I'd read, is to use a diode on the
complementary NPN/PNP side (and a resistor in parallel, I
gather.) Supposedly, it flatten out the gain curve in just
the right amount to balance things pretty well. It's an
interesting point, if true, because the quasi-complementary
output stage is attractive in that it can use the exact same
NPN part number for both quadrants' output BJTs.

I need to post up some different output structures in the
wane hope someone will help me walk through an analysis of
them.

This article states that a Baxandall diode made little change in linearity:
http://www.embedded.com/design/206801065?printable=true

Thanks for the article. Same author!! If I read closely
enough what he is saying, he is saying that the Baxandall
diode adds little _in the case of class-A operation_. In the
case of class-B, I think he argues it is worth having!

He writes,

"The choice of class A output topology is now simple.
For best performance, use the CFP. Apart from greater
basic linearity, the effects of output device
temperature on Iq are servoed out by local feedback,
as in class B. For utmost economy, use the quasi
complementary with two NPN devices: these need only a
low Vce(max) for a typical class A amp, so here is an
opportunity to recoup some of the money spent on
heatsinking.

"The rules are different from class B; the simple quasi
configuration will give first class results with
moderate NFB, and adding a Baxandall diode to simulate
a complementary emitter follower stage makes little
difference to linearity."

I think I represented the meaning of these two paragraphs,
accurately. And if you look closely at Figure 4, you will
see that there are two curves -- both are for quasi-
complementary outputs. However, one of them is class-B --
the really nasty-looking one. That one cries out for fixing
and is exactly what I was just talking about, above!!! So
the Baxandall diode really seems to be useful in allowing one
to _select_ class-B operation without having to pay much for
it. Makes class-B lots more attractive with quasi-
complementary outputs.

Let me know if you think otherwise.

There is also a Baxandall Tone Control circuit which is discussed here:
http://digitalcommons.calpoly.edu/eesp/14/

I'd need to download that thing to read it, I guess. For
now, I merely suspect that Baxandall writes about ideas from
time to time and isn't known for one thing.

For my own purposes, distortion of anything less than about 1% is probably
not worth paying for or striving to achieve. For audio, my ears are not all
that good and I would probably welcome distortion in the form of non-linear
frequency response to compensate for degraded sensitivity at the high end.

I don't have a number in mind because I'm _very_ ignorant
about what I'd care about and what I wouldn't. I _do_ know
one thing.... I really _hate_ the 10% THD computer speaker
systems. That much I do know. There is a place in hell for
people who pawn those things off as amplifiers with a nickel.

And there is also the argument that any sound that is naturally produced
will have some significant distortion that is actually part of the
listener's experience.

hehe. I'm imagining Mister Magoo right now and what he'd
consider "good." With Magoo as the "listener" ....
Alfred E Neuman's "What, me worry?" comes to mind regarding
any amplifier system.

Slightly more seriously, best of all would be that we somehow
analyze each and every person's brain's responses to sound,
in real time if possible, from the conscious interpretation
back through to the cochlea and the transducers nearby, to
the environment around it, and use a DSP to process the
content first before driving a speaker system, at all.

I expect to be dead before that happens, though.

I'm good ignoring "listener's experience" and focusing on a
more objective measure of some kind, letting the chips fall.

Whatever the acoustics in a given auditorium may be,
they contribute to the waveshape as it is received by the listener's ears,
and it varies depending on where one is seated. Sometimes added distortion,
such as an echo, may enhance the enjoyment of the music, and coloration due
to an imperfect amplifier might just as easily be perceived as pleasant
rather than objectionable. It is in fact distortion that causes an
audiophool to prefer the "warm" sound of a tube amplifier over a laboratory
grade solid state amplifier.

Those arguments are "beyond my pay grade." I'll just retreat
to something I can actually compute.

I am more impressed with amplifiers that are extremely efficient, such as
PWM amps. And for some types of test equipment that I have designed, I had
to deal with maintaining phase shift to better than 1 degree into a range
of inductive, resistive, or capacitive loads, and with outputs of power
line frequencies of 45-450 Hz, for voltage sources up to 300 VAC, and
current sources up to 100 amperes, at 50 VA to 300 VA or higher. And they
had to be able to withstand overloads and short circuits.

I like efficiency as one goal. Especially as I'm getting to
understand just how much power can be wasted without much
value. This 10W thing, if we are talking about class-A and
planning for 6db overhead (4X) as someone I read in one of
those articles saying about it, might mean a 40W capability
into 8 ohms, rails that are way out there and power waste
that starts to look like a toaster. So I'm beginning to get
my head turned 'round even at 10W!! Cripes, that spec is
rapidly becoming something I'm beginning to respect a lot
more and to realize that I might have landed on a number that
is better for teaching than I'd first imagined it to be.

Jon

 

[David Eather]

Jon Kirwan wrote:

On Sun, 31 Jan 2010 12:38:49 -0800 (PST), David Eather
eather@tpg.com.au> wrote:

On Jan 30, 9:06 am, Jon Kirwan <j...@infinitefactors.org> wrote:
On Fri, 29 Jan 2010 13:49:16 -0800 (PST), David Eather
snip

Yes and No. All the published circuits are made by people who want to
sell transistors,
A concern I care not the least about. My _real_ preference,
were I to impose it on the design, would be to use ONLY
PN2222A BJTs for all the active devices. One part. That's
it. Why? Because I've got thousands of them.

Literally. Something like 22,000 of the bastards. I give
them away like popcorn to students at schools. Got them
_very cheaply_. So if I were pushing something, I'd be
pushing a 10W PN2222A design, use signal splitting approach
probably (because it's the only way I think think of, right
now), and distribute the dissipation across lots and lots of
the things.

What to go there?
Signal Splitting? Can you sketch out what your thinking?

Yeah, I think so. Something like this:

: | |
: \ |
: / R2 |
: \ |
: / |
: | |
: | |/c Q2
: +---------|
: | |>e
: | |
: |/c Q3 |
: -------| +-----
: |>e |
: | |
: | |/c Q1
: +---------|
: | |>e
: | |
: \ |
: / R1 |
: \ |
: / |
: | |

The "signal splitter" here is Q3. It's also providing gain,
too, though. The emitter and collector move in opposite
directions and the signal "splits" at Q3. (The emitter
follows the base, the collector inverts the base.)

If I read with any understanding about these things, properly
biasing Q3 is a pain, the Q3 gain varies with the load itself
as well as its bias, and compensation issues are complicated
a bit.

The wikipedia type circuit can use a few n2222 - I count a max of 5.
Even if you could use only n2222 it would not be a good idea - making
the circuit stable would be more difficult.

Yes, ignorant as I am still of the details, I think that's
very true. The splitter has significant signal voltage on
its input and I've read that pole-splitting methods for
improving stability are harder to apply here.

On the good side the n2222
is a good choice for Q1,Q2 and as active replacements for R5,R6 and
one other (optional) we haven't met yet. What makes it a good
transistor is the large current gain / bandwidth product and the flat
DC current gain over a wide range of viable bias currents. Both
contribute to low distortion.

http://www.onsemi.com/pub_link/Collateral/P2N2222A-D.PDF (page 3
graph)

compared to say 2n3904
http://www.onsemi.com/pub_link/Collateral/2N3903-D.PDF

where the flat portion of the DC gain curve is over a very limited
range.

Interesting point to consider. Something that had slipped by
me, so far.

not audio systems, power supplies or transformers.
Got it.

As a result the power supply is often assumed to be regulated, which
is not true in this case, or the power supply is treated in a very
perfunctory manner that is not at all compatible with good design.
In this case you have the voltage you need for the 10 watts, plus
voltage drop for the driver circuitry and output stage , plus ripple
voltage, plus whatever is required for transformer regulation and
mains regulation. When you add it all up you might find that a chosen
transistor/component is actually not at all suitable for the job. Back
to the drawing board. Change this change that recheck everything again
etc.
In this case, though, there is nothing particularly
remarkable about the rails. Taken across the entire span,
even, doesn't exceed the maximum Vce of a great many BJTs. So
no real worry there. But I see some of where problems may
arise. Luckily, at this level I can side-step worrying about
that part and get back to learning about amplifier design,
yes?
I come up with a figure of 50 volts rail to rail no load voltage -
after picking out a common transformer with 15% regulation.

A 30 volt CT transformer with 15% regulation and 7% mains over-voltage,
less voltage drop for the diode bride would give rails of +/- 25.

A dual 12.6 volt transformer would give a minimum (worst case with
transformer at full load and mains 7% under voltage) of 16.something
volts meaning big filter caps if you were serious at getting 10 watts.
One of the reasons to go PSU first I think. (Also I live in a tiny
jerk-water town where no one knows what a custom made transformer is let
alone where you can get one wound)

Okay. This is going to force me to sit down with paper and
work through. I was stupidly imagining +/-18V max, or 36V
rail to rail. I haven't considered the details of the output
section yet, driving a load from rails that run up and down
on capacitors that charge and discharge at 1A-level currents
into the load, and perhaps I need to spend some more time
there before moving on.

There are so many ways to cut this. Start at the input and
that's one focus that may work okay. Start at the output
stage and that provides important power supply information,
though. So maybe I should start at that end?

If you do the power supply first you have the figures needed for your
worst case already. It saves time and makes a better result (no
tendency to comprimise to save all the calculations already done).
Well, does this mean we should hack out the power supply
first? I'm perfectly fine with that and can get back to you
with a suggested circuit and parts list if you want to start
there. We could settle that part before going anywhere else
and I'd be happy with that approach, too, because to be
honest I don't imagine it to put a horrible delay into
getting back to amplifier design. So I'm good either way.
I'm looking at some of your other posts and I don't think you need a
maths lesson from me. If you want to do a power supply great. Its a
small one so nothing much too it. If you don't want, I'm OK too.

I still haven't been down the path on my own, yet. So I
don't have strong opinions about this. It's like going to
Disneyland for the first time. Which land should I go to,
first? Later, after being there a few times, I may look at
the flow of people and decide that "Adventureland" is the
best first start. But first time out? Who knows? I'm open
to guidance. Everything is new.

I should also ask if you have a multi meter, oscilloscope (not necessary
but useful)and how is your soldering? But it would be wise to keep this
whole thing as a paper exercise before you commit to anything.
I have a 6 1/2 digit HP multimeter, a Tek DMM916 true RMS
handheld, two oscilloscopes (TEK 2245 with voltmeter option
and an HP 54645D), three triple-output power supplies with
two of them GPIB drivable, the usual not-too-expensive signal
generator, and a fair bunch of other stuff on the shelves.
Lots of probes, clips, and so on. For soldering, I'm limited
to a Weller WTCPT and some 0.4mm round, 0.8mm spade, and
somewhat wider spade tips in the 1.5mm area. I have tubs and
jars of various types of fluxes, as well, and wire wrap tools
and wire wrap wire, as well. I also have a room set aside
for this kind of stuff, when I get time to play.
OK. Next serious project, I'm coming around to your place!
You come to the west coast of the US and I'll have a room for
you!
Your gear is
better than mine. I had to ask, rather than just assume just in case my
assumptions got you building something you didn't want to, and got you
splattered all over the place from the mains, or suggesting you choose
the miller cap by watching the phase shift of the feedback circuit - I
don't read a lot of the posts so I didn't know what you could do.
To be honest, I can do a few things but I'm really not very
practiced. My oscilloscope knowledge is lacking in some
areas -- which becomes all too painfully obvious to me when I
watch a pro using my equipment. And I'm still learning to
solder better. It's one of a few hobbies.
Jon
Have a look at
http://en.wikipedia.org/wiki/Electronic_amplifier
Done.
The bits on class A might be interesting as it says 25% efficiency and
50% obtainable with inductive output coupling (i.e. with a transformer)
which is what I said, not what blow hard Phil said.
What I first see there is the amplifier sketch at the top of
the page
I wasn't going to prompt, but it is close to the sort of thing, I
think, you should be aiming for . As someone has already noted (I
would attribute you if I wasn't on GG, I'm sorry) it has been drawn up
for a single supply, rather than a more common (for this size /
configuration) split supply.
I had assumed we'd be using a split supply.

I think that's very much the preferred way.

I feel more comfortable assuming it, too.

I had assumed a speaker would be hooked up via a cap to the
output, so DC currents into a speaker coil would be removed
from any concern. But I was also holding in the back of my
mind the idea of tweaking out DC bias via the speaker and
removing the coupling cap as an experiment to try. And if
so, I'd pretty much want the ground as a "third rail."
Exactly right! There are two common ways to reduce/remove any offset
from the output. Neither is shown on the wikipedia circuit. If you
have another split rail circuit it will certainly have one method -
both methods involved use the diff amp.

Thanks.

(Playing just a bit upon the Chicago parlance about the once
dangerous rail in their transit system.)

(I don't really care too much about arguing about
efficiencies right now -- I'm more concerned about learning.)
The input stage shown is a voltage-in, current-out bog
standard diff-pair. First thing I remember about is that R4
shouldn't be there
Correct. Theory says it does nothing. I practice the theory but have
the occasional heretical belief about that.
Actually, I think I've read that theory says it is _better_
to be removed. The reason seemed pretty basic, as it's
easier to get close to a balanced current split; and this, I
gather, lowers 2nd harmonic distortions produced in the pair
-- notable more on the high frequency end I suppose because
gain used for linearizing feedback up there is diminishing
and can't compensate it.

In other words, it's not neutral. It's considered to be
better if I gathered the details. Then even better, the
current mirror enforces the whole deal and you've got about
the best to be had.

Of course, mostly just being a reader means I have no idea
which end is up. So I might have all this wrong.
No. Thats all correct. I'll show a different circuit latter

Okay. I'll enjoy the moment when it happens.

and better still both R3 and R4 should be
replaced with a current mirror.
This would provide more differential gain.
_and_ improve distortion because the currents are forced to
be balanced in the pair, yes?
yes.

Okay. So I am picking up details not too poorly, so far.
(Seriously, I think you are doing amazing)

R5 should be a replaced with
a BJT, as well.
In the right configuration it would reduce the common mode signal gain
of things like mains hum and supply ripple (you mentioned power supply
isolation before).
Yes, that's how I thought about it.

Also, from another (what do you call it branch? thread?) you were
discussing boot-strapping R6. This is not done so much as amplifiers
get bigger but a BJT configured in the same way as the replacement for
R5 is very common. I'm leaving the details to you - perhaps there is a
way to reduce component count without affecting performance. (I am
hoping this is what you wanted "nutting it out for yourself")
Yes! I don't want things handed on a platter. But I also
don't want to have to rediscover all of the ideas by making
all of the mistakes, either. This is the kind of "pointer"
towards something that I like a lot. It gives me a place to
think about something, but leaves me some reason to have to
do so and that helps me own it better.

One general truth about learning is that you don't present
someone with a problem so out of their depth that they have
no chance at it. Doing that means they fail, they feel like
a failure, and it causes a student to just want to go away.
They lose motivation, usually, in cases like that. On the
other hand, providing no difficulty at all merely means
repetition of what they already know and they grow bored from
that, too. Finding the sweet spot where a student is faced
with interesting problems that are not already known, but
perhaps within reach of grasping at with some effort, is the
key. Then it can be fun, educational, and motivate.

That's what you just did for me.

I assume the input impedance of that example
is basically the parallel resistance of R1 and R2, but if we
Yes.
Okay.
There is the parallel resistance of R5 x Beta Q1 as well, but this is
normally so high it won't affect the result. And if R5 is replaced
with an active device it can become essentially infinite.

Okay. I've got that detail from other discussions, too. So
yes, understood. Also, I mentioned replacing R5, I think. In
replacing R5 with active parts, I'm thinking of two BJTs in a
usual form that seems to work pretty well over supply
variations.

Two BJT's make a current sink with a nice sharp knee giving best CMRR.
I'll point out a small "optimization" (price/component reduction but
with a small degradation in performance - given the over abundance of
2n2222 in your area it probably doesn't count). You could use a single
BJT current sink with a voltage reference (2 diodes, low voltage zenner
or LED) tied to the negative rail. You could then use the same reference
voltage for a 1 BJT current sink in place of R6. Saving a couple of
components.

use split supplies I'd imagine replacing the two of them with
a single resistor to the center-ground point.
Yes, but you should probably think of a whole passive network to
filter out low and high frequency - (think what happens if you amp is
operated near a source of RF)
Well, every trace picks up like little antennae. All kinds
of trace voltages appearing here and there. Not good.

So. Can you make an audio amplifier that can withstand a
microwave oven environment and deliver good performance while
irradiated with 1kW banging around in there?
If you can do that the military wants you to EMP harden all there
electronics. The input is a little different because some user always
want to stick a bloody gret big long wire onto it.



I actually _do_ work on low-mass, direct-contact temperature
measuring devices designed to work within a microwave
environment. (But no electronics or metals inside.)

But you brought up the microwave environment, so I hope you
don't mind the teasing about it.

No.

There's no
miller cap on Q3,
Depending on transistors layout etc it might not be needed, but more
often it is the size that is the question.
I was thinking it helped locally linearize the VAS section
and that such would be "good" most anywhere. But I am just
taking things without having worked through them on my own.
So...
It sets the bandwidth of the VAS stage so you can use negative
feedback without the whole thing turning into smoke. Do you know of
control theory / bode diagrams. There is a minuscule amount needed
for this app.

I am familiar with _some_ closed loop control theory,
sufficient to get me by with PID controls (using _and_
writing code for them.) Bode diagrams are something I have
not used, though I've seen them. My math is adequate, I
suspect. But I will have to read up on them, I suppose.

For Laplace analysis, I'm familiar with complex numbers,
poles and zeros, partial fraction extractions, and so on.
Just inexperienced in the "short cuts" that many use to get
(and think about) answers.

Bode diagrams are really simple and (for this type of thing) will get
you where you want to go.

I'd probably replace the two diodes with
one of those BJT and a few resistor constructions I can't
remember the name of (which allows me to adjust the drop.)
Vbe multiplier...
Okay. Thanks.

The feedback ... well, I need to think about that a little
more. There's no degen resistors in the emitters of Q4 and
Q5.
Why would/should you use them?

Jon,
Just in case the question misled you - I was asking only about things
that needed changing. So I asked a question about the degen Rs on Q4 and
Q5, even though you would/should use them and for a few good of reasons
of which I think thermal stabilization is the more important.

I'm still thinking about that. In general, I was thinking
about them because of the "little re" that is kT/q based in
each BJT, and varies on Ie. Since Ie is varying around, I
was thinking about something fixed there to overwhelm it and
"make it knowable" for the design, I suppose. Maybe that's
all wet, given your query. I'll toss the idea off the side,
for now.
Try working through the thermal stabilization. Just make a stab at the
transistor junction temperatures - it will be pretty hot (unless you
can afford mega bucks for heatsinking)

I need to understand the output configuration a little better
before I do that.

Including thinking more closely about swinging one end of an
output cap around so that 1Amp rms can pass through it at
20Hz. I = C dv/dt, but V=V0*sin(w*t), so I=C*w*V0*cos(w*t).
Assuming max current at the max slew rate for a sine at phase
angle zero, the w*t is some 2*PI*N thing, so cos(w*t) goes to
1. That makes I=w*C*V0. But w=2*pi*20, or about 126 or so.
So I=126*C*V0. So with I=1A, C=1/(126*V0). With V0=15V, I
get about 530uF for the output cap. That's an amp peak only
at the right phase, too. It'll be less elsewhere. To make
that an amp rms, the cap would need to be still bigger.

Peak current via the cap will take place right about the time
when the two BJTs's emitters are at their midpoint. One of
the BJTs will be supplying that. Not only that, but also
depending upon class mode of operation, supplying current to
the other one as well. How much is important to figuring out
the wattage.

I need to sit down with paper, I suspect. But if you want to
provide some suggested thinking process here, I'd also be
very open to that, as well. I'll take a shot at it either
way, but it helps to see your thinking, too. If you can
afford the moment for me.

If I had a split power supply I would *always* get rid of the output
capacitor. It is not difficult to get the output DC to withing 50mv of
gnd. A weird thing I have noticed, and I think you would have noticed it
sooner, is that no one, even audio "golden ears" pay serious attention
to the output cap. They just stick a plain old electrolytic of no
particular type (some times it's a bipolar) in the output, make it
bigger than needed for the LF -3db corner and call it "good". It would
seem that some attention should be paid to "ripple" current at
frequencies like 20khz etc, so some low esr caps would seem mandatory.
That music has relatively less high frequency components is the only
reason I can think of that this very lax approach might work.

Um.. okay, I need to sit down and think. Mind is spinning,
but I've not set a finger to paper yet and there is lots to
think about in that one. I could be way, way off base.
Not at all.
Thanks for that. I'm just glad to be able to talk to someone
about any of this, at all. So please accept my thanks for
the moments you are offering.

Is there a way you could post a schematic of where your thinking is
and what you would like to discuss - there is no need for a complete
circuit.
Yes. I can use ASCII here, for example. But before I go off
into the wild blue with this, do you want to focus on the
power supply first? Or just jump in on the amplifier?
I don't mind. Earlier I put a stab at a no load worst case voltage,
you can use that if you want to. Until you get to output stage power
dissipation that is all you need.

Maybe I'd like to focus on understanding different output
pair configurations, first. I frankly don't like the "haul
the output pair around with a collector on one side and a
resistor on the other with a rubber diode in between to keep
them biased up" approach. It's smacks of heavy-handedness
and I simply don't like the way it looks to me. Everything
tells me this works, but it is indelicate at the very least.

Well it would be a Darlington or complementary pair (I can't remember
the name sz...? )for the output transistors but I have no objections

However, it is crucial that I understand it in detail before
deciding what I really think about it. For example, I might
want to replace the resistor with a current source.

Sorry you lost me on which R

But

without apprehending the output stage more fully, its time
domain behavior over a single cycle for example, I'm not
comfortable with hacking it here and there, ignorantly.

Jon

 

[pimpom]

David Eather wrote:

If I had a split power supply I would *always* get rid of the
output
capacitor. It is not difficult to get the output DC to withing
50mv of
gnd. A weird thing I have noticed, and I think you would have
noticed
it sooner, is that no one, even audio "golden ears" pay serious
attention to the output cap. They just stick a plain old
electrolytic
of no particular type (some times it's a bipolar) in the
output, make
it bigger than needed for the LF -3db corner and call it
"good". It
would seem that some attention should be paid to "ripple"
current at
frequencies like 20khz etc, so some low esr caps would seem
mandatory.
That music has relatively less high frequency components is the
only
reason I can think of that this very lax approach might work.

If I may inject a comment here: I strongly support the idea of
avoiding an output coupling capacitor. I always use a split-PS,
OCL configuration unless some other consideration makes it
necessary to use a single-ended PS.

The comment about DC offset at the output terminal reminds me of
an experience I had more than 20 years ago. I was asked to spruce
up the P.A system at our state legislators' main session hall.
One of the things I did was to replace the old tube power amp
with my own design. I built four 60-watt amps (3 in use, one
spare) using 2N3055 BJTs in quasi-complementary configuration (I
couldn't easily get true complementary pairs then). Since the
existing system distributed audio power to dozens of small
speakers, inside and outside the hall, over a standard 100-volt
line, I integrated a 4-ohm input, 100V output transformer in my
amps.

When I first tested the system, one output transistor each in two
of the amplifiers warmed up quickly even without any output - not
actually hot, but warmer than they should be. After a few moments
of puzzlement, I traced the culprit to slight DC offset at the
output terminal. It was only a small fraction of a volt and
wouldn't have mattered with direct coupling to a speaker. But the
DC resistance of the primary winding of the output transformer
was so low (a fraction of an ohm) that it forced one of the
output transistors to draw a substantil amount of DC current at
idle.

I further traced the cause of offset to poorly matched
transistors at the input differential stage. I didn't include
provision for manual balancing of the static DC level, so I tried
out a few transistors for the input stage until I got a pair that
matched closely enough to reduce the offset to within a millivolt
or so (there was no hope of obtaining a factory-matched pair).

I know this has no direct relevance to the discussion, but I was
partly reminiscing and partly thinking that it may not be a bad
idea to give a real-life example of how easy it is to overlook
something.

 

[Paul E. Schoen]

"pimpom" <pimpom@invalid.invalid> wrote in message
news:hkcqg6$vda$1@news.albasani.net...

David Eather wrote:

If I had a split power supply I would *always* get rid of the output
capacitor. It is not difficult to get the output DC to withing 50mv of
gnd. A weird thing I have noticed, and I think you would have noticed
it sooner, is that no one, even audio "golden ears" pay serious
attention to the output cap. They just stick a plain old electrolytic
of no particular type (some times it's a bipolar) in the output, make
it bigger than needed for the LF -3db corner and call it "good". It
would seem that some attention should be paid to "ripple" current at
frequencies like 20khz etc, so some low esr caps would seem mandatory.
That music has relatively less high frequency components is the only
reason I can think of that this very lax approach might work.


If I may inject a comment here: I strongly support the idea of avoiding
an output coupling capacitor. I always use a split-PS, OCL configuration
unless some other consideration makes it necessary to use a single-ended
PS.

The comment about DC offset at the output terminal reminds me of an
experience I had more than 20 years ago. I was asked to spruce up the P.A
system at our state legislators' main session hall. One of the things I
did was to replace the old tube power amp with my own design. I built
four 60-watt amps (3 in use, one spare) using 2N3055 BJTs in
quasi-complementary configuration (I couldn't easily get true
complementary pairs then). Since the existing system distributed audio
power to dozens of small speakers, inside and outside the hall, over a
standard 100-volt line, I integrated a 4-ohm input, 100V output
transformer in my amps.

When I first tested the system, one output transistor each in two of the
amplifiers warmed up quickly even without any output - not actually hot,
but warmer than they should be. After a few moments of puzzlement, I
traced the culprit to slight DC offset at the output terminal. It was
only a small fraction of a volt and wouldn't have mattered with direct
coupling to a speaker. But the DC resistance of the primary winding of
the output transformer was so low (a fraction of an ohm) that it forced
one of the output transistors to draw a substantil amount of DC current
at idle.

I further traced the cause of offset to poorly matched transistors at the
input differential stage. I didn't include provision for manual balancing
of the static DC level, so I tried out a few transistors for the input
stage until I got a pair that matched closely enough to reduce the offset
to within a millivolt or so (there was no hope of obtaining a
factory-matched pair).

I know this has no direct relevance to the discussion, but I was partly
reminiscing and partly thinking that it may not be a bad idea to give a
real-life example of how easy it is to overlook something.

Something else to consider is a bridge output connection. You just need two
output stages, invert the phase to one, and make sure the DC voltages on
each are balanced. Another bonus is that you can get nearly 24 volts P-P
with a 24 VDC single supply. You can get very close to the rails if the
driver stage uses a slightly higher power supply voltage, so you can
optimize efficiency but at the cost of distortion (clipping).

At one time I considered making an amplifier with a dynamically adjustable
power supply so that the rails would always be just a couple of volts above
the peak output signal. It would probably be workable for a high-power
signal generator where some clipping can be tolerated as the output is
increased, but for music or other complex signals that vary unpredictably
in amplitude, there would need to be a delay in the signal long enough to
allow the power supply to adjust to what will be needed. At low
frequencies, it might be possible even to have the power supply track the
waveshape and even higher efficiency could be obtained.

I have used serial analog delay lines which are basically a bucket brigade
of switched capacitors, clocked higher than the maximum frequency required.
I designed a phase-shifting circuit for power line frequency, using an IC
that was sold at Radio Shack at the time, an SAD1024
http://www.geofex.com/sad1024.htm. I think I clocked it at a rate which
produced a 90 degree phase shift at 60 Hz, which would be 1024/0.00416 =
246 KHz. But it was prone to distortion, and it was not long before the IC
was discontinued and replaced with a dual 512 stage device that was even
worse.

There are much better ways to accomplish such feats now, but in 1980 or so
there were not many alternatives. Now the way to do it might be to digitize
the signal and then use a circular buffer to achieve whatever delay is
needed. Probably 16 bit audio sampled at 44 kHz so you can get about a 1.5
second delay with a 16 bit x 64k word memory. But with all that, probably a
PWM amp would be the way to go.

Just draining the brain through my fingers and the keyboard into
cyberspace...

Paul

 

[David Eather]

Paul E. Schoen wrote:

"pimpom" <pimpom@invalid.invalid> wrote in message
news:hkcqg6$vda$1@news.albasani.net...
David Eather wrote:
If I had a split power supply I would *always* get rid of the output
capacitor. It is not difficult to get the output DC to withing 50mv of
gnd. A weird thing I have noticed, and I think you would have noticed
it sooner, is that no one, even audio "golden ears" pay serious
attention to the output cap. They just stick a plain old electrolytic
of no particular type (some times it's a bipolar) in the output, make
it bigger than needed for the LF -3db corner and call it "good". It
would seem that some attention should be paid to "ripple" current at
frequencies like 20khz etc, so some low esr caps would seem mandatory.
That music has relatively less high frequency components is the only
reason I can think of that this very lax approach might work.

If I may inject a comment here: I strongly support the idea of avoiding
an output coupling capacitor. I always use a split-PS, OCL configuration
unless some other consideration makes it necessary to use a single-ended
PS.

The comment about DC offset at the output terminal reminds me of an
experience I had more than 20 years ago. I was asked to spruce up the P.A
system at our state legislators' main session hall. One of the things I
did was to replace the old tube power amp with my own design. I built
four 60-watt amps (3 in use, one spare) using 2N3055 BJTs in
quasi-complementary configuration (I couldn't easily get true
complementary pairs then). Since the existing system distributed audio
power to dozens of small speakers, inside and outside the hall, over a
standard 100-volt line, I integrated a 4-ohm input, 100V output
transformer in my amps.

When I first tested the system, one output transistor each in two of the
amplifiers warmed up quickly even without any output - not actually hot,
but warmer than they should be. After a few moments of puzzlement, I
traced the culprit to slight DC offset at the output terminal. It was
only a small fraction of a volt and wouldn't have mattered with direct
coupling to a speaker. But the DC resistance of the primary winding of
the output transformer was so low (a fraction of an ohm) that it forced
one of the output transistors to draw a substantil amount of DC current
at idle.

I further traced the cause of offset to poorly matched transistors at the
input differential stage. I didn't include provision for manual balancing
of the static DC level, so I tried out a few transistors for the input
stage until I got a pair that matched closely enough to reduce the offset
to within a millivolt or so (there was no hope of obtaining a
factory-matched pair).

I know this has no direct relevance to the discussion, but I was partly
reminiscing and partly thinking that it may not be a bad idea to give a
real-life example of how easy it is to overlook something.

Something else to consider is a bridge output connection. You just need two
output stages, invert the phase to one, and make sure the DC voltages on
each are balanced. Another bonus is that you can get nearly 24 volts P-P
with a 24 VDC single supply. You can get very close to the rails if the
driver stage uses a slightly higher power supply voltage, so you can
optimize efficiency but at the cost of distortion (clipping).

At one time I considered making an amplifier with a dynamically adjustable
power supply so that the rails would always be just a couple of volts above
the peak output signal. It would probably be workable for a high-power
signal generator where some clipping can be tolerated as the output is
increased, but for music or other complex signals that vary unpredictably
in amplitude, there would need to be a delay in the signal long enough to
allow the power supply to adjust to what will be needed. At low
frequencies, it might be possible even to have the power supply track the
waveshape and even higher efficiency could be obtained.

I have used serial analog delay lines which are basically a bucket brigade
of switched capacitors, clocked higher than the maximum frequency required.
I designed a phase-shifting circuit for power line frequency, using an IC
that was sold at Radio Shack at the time, an SAD1024
http://www.geofex.com/sad1024.htm. I think I clocked it at a rate which
produced a 90 degree phase shift at 60 Hz, which would be 1024/0.00416 =
246 KHz. But it was prone to distortion, and it was not long before the IC
was discontinued and replaced with a dual 512 stage device that was even
worse.

There are much better ways to accomplish such feats now, but in 1980 or so
there were not many alternatives. Now the way to do it might be to digitize
the signal and then use a circular buffer to achieve whatever delay is
needed. Probably 16 bit audio sampled at 44 kHz so you can get about a 1.5
second delay with a 16 bit x 64k word memory. But with all that, probably a
PWM amp would be the way to go.

Just draining the brain through my fingers and the keyboard into
cyberspace...

Paul

The adjustable power supply approach has been dubbed type "H", but like
type "G", it is almost certainly dying now... (which is sad I think - it
does cut down lines of investigation for originality and inspiration)

In the search for optimal efficiency:

http://www.irf.com/product-info/datasheets/data/irs2092.pdf

(irf - the worlds most experimenter unfriendly company)

 

[Paul E. Schoen]

"David Eather" <eather@tpg.com.au> wrote in message
news:_ZidnV6rPbIxhffWnZ2dnUVZ_rednZ2d@supernews.com...

The adjustable power supply approach has been dubbed type "H", but like
type "G", it is almost certainly dying now... (which is sad I think - it
does cut down lines of investigation for originality and inspiration)

In the search for optimal efficiency:

http://www.irf.com/product-info/datasheets/data/irs2092.pdf

(irf - the worlds most experimenter unfriendly company)

Here's what I found for Class H, which used a power supply boost under
certain high output conditions:
http://www.nxp.com/documents/data_sheet/TDA1562Q_ST_SD.pdf

But it has been discontinued, and . I didn't realize that it had already
been implemented and had a class letter. I came up with the idea in the
early 80s IIRC. And I also tried to design a switching amplifier a few
years later. Maybe they were novel ideas then and I should have pursued
them.

My idea for the switching amplifier planned to use, basically, two
programmable switching supplies driven by the input signal and its inverse.
But it ignored the necessity of supplying both positive and negative
current. The answer, of course, was an H-bridge configuration. And I think
sometimes that is known as Type H, as opposed to Class H.

Another engineer at the time thought that he could just pass a pulse-width
modulated high frequency signal through a ferrite transformer rated at
perhaps 200 VA and 40 kHz. It would be modulated by the 60 Hz nominal
signal that was to be provided as an output, and would use capacitors and
inductors to filter out the carrier. But the only way for this to work at
all would be to rectify the output, resulting in half of the waveform. And
that would also saturate the core. So it was doomed from the start, but he
actually had transformers made and PC boards built before he could be
convinced that it had a fatal error.

That IRF part sounds like a real beast. There's pretty much no need to
design a power amplifier from discrete parts if you can get that IC for
about $5 and some $2 MOSFETs to make a reliable, rugged, and efficient
amplifier. Their development kit is $200 but that's not bad for a 250W x 2
channel amp.

I guess that's what you mean about them being experimenter unfriendly. No
reason to design and breadboard and fiddle around with something if it's
already been done essentially to perfection..

Paul

 

[David Eather]

Paul E. Schoen wrote:

"David Eather" <eather@tpg.com.au> wrote in message
news:_ZidnV6rPbIxhffWnZ2dnUVZ_rednZ2d@supernews.com...
The adjustable power supply approach has been dubbed type "H", but like
type "G", it is almost certainly dying now... (which is sad I think - it
does cut down lines of investigation for originality and inspiration)

In the search for optimal efficiency:

http://www.irf.com/product-info/datasheets/data/irs2092.pdf

(irf - the worlds most experimenter unfriendly company)

Here's what I found for Class H, which used a power supply boost under
certain high output conditions:
http://www.nxp.com/documents/data_sheet/TDA1562Q_ST_SD.pdf

But it has been discontinued, and . I didn't realize that it had already
been implemented and had a class letter. I came up with the idea in the
early 80s IIRC. And I also tried to design a switching amplifier a few
years later. Maybe they were novel ideas then and I should have pursued
them.

My idea for the switching amplifier planned to use, basically, two
programmable switching supplies driven by the input signal and its inverse.
But it ignored the necessity of supplying both positive and negative
current. The answer, of course, was an H-bridge configuration. And I think
sometimes that is known as Type H, as opposed to Class H.

Another engineer at the time thought that he could just pass a pulse-width
modulated high frequency signal through a ferrite transformer rated at
perhaps 200 VA and 40 kHz. It would be modulated by the 60 Hz nominal
signal that was to be provided as an output, and would use capacitors and
inductors to filter out the carrier. But the only way for this to work at
all would be to rectify the output, resulting in half of the waveform. And
that would also saturate the core. So it was doomed from the start, but he
actually had transformers made and PC boards built before he could be
convinced that it had a fatal error.

That IRF part sounds like a real beast. There's pretty much no need to
design a power amplifier from discrete parts if you can get that IC for
about $5 and some $2 MOSFETs to make a reliable, rugged, and efficient
amplifier. Their development kit is $200 but that's not bad for a 250W x 2
channel amp.

I guess that's what you mean about them being experimenter unfriendly. No
reason to design and breadboard and fiddle around with something if it's
already been done essentially to perfection..

Paul


No, I mean they are pricks to try and deal with - unless you have a very

large budget.

 

[Jon Kirwan]

On Thu, 04 Feb 2010 05:46:33 +1000, David Eather
<eather@tpg.com.au> wrote:

A 30 volt CT transformer with 15% regulation and 7% mains over-voltage,
less voltage drop for the diode bride would give rails of +/- 25.

I'm not entirely familiar with the formal terms, but I figure
the 15% regulation you mentioned above must mean that the
ripple voltage goes from 100% to 85%. In general, the
equation for the angle would be something like:

angle = arcsin( 1 - Rf*(1-Vd/Vpk) )

This is with Rf being the ripple factor (not in percent
terms, obviously) and Vd being the sum of the diode drops
(full wave would be something like 2V) and Vpk being the
sqrt(2)*Vrms. At least, that's what the equation works out
for me on paper. (I can develop out here, if needed.)

The peak diode current happens just as the first moment of
conduction (which is neatly defined by the ripple factor, if
I understand you) and would be something like:

Ipk = 2*pi*f*C*Vpk*cos( angle )

Since the cos(arcsin(x)) is just sqrt(1-x^2), the computation
looks like:

Ipk = 2*pi*f*C*Vpk*SQRT(1-(1-Rf*(1-Vd/Vpk))^2)

There's probably some other adjustments to nail it, but that
probably gets somewhat close.

For a full wave bridge (I know, I think you were talking
about a half wave, but let's go with this for a moment) with
Vpk=25.2*SQRT(2) and Vd=2V (for two diodes in conduction in
the bridge) and Rf=.15 and f=60 (US-centric) and let's say a
C=2200uF, that gets something like Ipk = 15.2A.

None of this takes into account the average load current or
peak load currents. It just assumes that the ripple factor
is somehow known to be correct. I started out assuming that
the average load current would be defined entirely by the
droop (which is, of course, based upon the ripple factor
assumption) and the time from the peak of a previous cycle to
the point at which the above angle occurs and conduction
again begins. But it is complicated slightly by the fact
that the transformer supplies the entire current draw (if it
can) for a short part of the cycle _after_ the peak, as its
slope is less than the droop slope of the cap. I played with
accounting for all that but then decided that in most
practical cases the droop is hopefully not too excessive and
if not, then the angle after the peak isn't that much.. maybe
5 to 8 degrees or so... So I decided to ignore it and just
rely upon the capacitor's droop only:

Iave = C*Vpk*Rf/((pi/2+angle)/(2*pi*f))

So in the case just mentioned, I get Iave = 1.7A. I take it
that Ipk can easily be a factor of 9 or 10 greater. Also, I
note that it would probably be helpful to have nifty charts
of some kind to help pick off details like these.

Can you expand a little on what you were talking about,
though? Was that a half wave suggestion? I'm not sure I can
make sense of the rails, if so. If not, then I'd still
appreciate some of the calculations so that I can sure I
follow all of it.

I'm guessing that if a rail is to have a minimum of 25V on it
at the bottom of the ripple, and you are talking about 15%
regulation, the peak is going to be 29.4V -- not counting the
diode drops. Add 1V for that and it's 30.4V. Another 7% on
top would be 32.7V for the peak. RMS would be that figure
divided by sqrt(2), wouldn't it? Or 23.1Vrms or so?

So would that suggest two of the cheap 25.2Vrms transformers
and plan on rails still slightly higher?

Oh, crap. The VA rating. That's another one to consider.
Later, I guess.

A dual 12.6 volt transformer would give a minimum (worst case with
transformer at full load and mains 7% under voltage) of 16.something
volts meaning big filter caps if you were serious at getting 10 watts.
One of the reasons to go PSU first I think. (Also I live in a tiny
jerk-water town where no one knows what a custom made transformer is let
alone where you can get one wound)

I suppose the old filament-type 12.6VAC transformers must be
common everywhere. I also see that Radio Shack (yes, I'm
holding my nose for a moment) still carries some "commodity"
type 25.2VAC CT 2.0A rated transformers for about US$10.
Their 12.6VAC CT 3.0A transformers are priced identically.

So what's considered to be generally available?

Jon

 

[Jon Kirwan]

There's another question that comes to mind regarding the
output stage. A lot of talk seems to revolve around
"crossover distortion." Seems almost very first thing folks
talk about when discussing class of operation if not also at
other times.

Seems to me that in a three-rail power supply situation
without an output capacitor involved, the crossover takes
place near the midpoint (ground) voltage between the rails,
at a time when current into the speaker load is also near
zero. (I'm neglecting any thoughts about inductance in the
speaker and physical coupling into the air, for now.) In
other words, where power at the speaker is near zero. Is it
really that important to consider?

I was looking at that terrible large scale gain plot for the
quasicomplementary output stage on the web site recently
mentioned in the thread (the lower curve in Figure 4 on this
link):

http://www.embedded.com/design/206801065?printable=true

(It's not that terrible of a plot, as the variation is from
..96 to .98 with the "normal" middle at .97.)

What's experience say here? Is it really so terrible as to
worry too much about something that takes place near zero
voltage, anyway? I'm just questioning the concern, for now.
I have no understanding about it, at all. Just wondering.

Jon

 

[Jon Kirwan]

On Thu, 04 Feb 2010 05:34:52 -0800, I wrote:

But it is complicated slightly by the fact
that the transformer supplies the entire current draw (if it
can) for a short part of the cycle _after_ the peak, as its
slope is less than the droop slope of the cap.

I overstated this. The capacitor does supply _some_ along
with the transformer windings during this short phase, as the
voltage on the cap is also declining with it.

Jon

 

[Paul E. Schoen]

"Jon Kirwan" <jonk@infinitefactors.org> wrote in message
news:upmjm55fpn755sdshfljjk6r7tdk9i6ug9@4ax.com...

On Thu, 04 Feb 2010 05:46:33 +1000, David Eather
eather@tpg.com.au> wrote:

A 30 volt CT transformer with 15% regulation and 7% mains over-voltage,
less voltage drop for the diode bride would give rails of +/- 25.

I'm not entirely familiar with the formal terms, but I figure
the 15% regulation you mentioned above must mean that the
ripple voltage goes from 100% to 85%. In general, the
equation for the angle would be something like:

Transformers are specified with a percentage regulation which means the
change in voltage from no load to full load conditions. A small,
inexpensive transformer might have 15% regulation so that the 30VCT unit
would have a 15% higher output voltage with no load, or 34.5 VRMS or 48.8
P-P. Mains voltage may vary +/- 7% or 120 VAC +/- 8, or 112 to 128 VAC. At
the high end of this range the tranny puts out about 52.2 V P-P. Assuming a
FWB rectifier and the CT as reference, with 0.7 V diode drop, you get 25.4
volts peak.

If you put a capacitor on the output, it eventually charges to the peak
voltage. This is the high limit that must be considered for design. It may
not be exact, and probably will be a bit lower, because a power transformer
is usually designed to operate in partial saturation, so the output will
not increase linearly above its design rating.

Under load, the output will drop, caused by the effects of primary and
secondary coil resistance as well as magnetic effects. These will cause
heating over a period of time, and the coil resistance will increase,
adding to the effect until a point of equilibrium is reached based on the
ambient conditions and removal of heat via conduction, convection, and
radiation.

Large power transformers, high quality audio transformers, and
instrumentation transformers are designed with perhaps 1% or 2% regulation,
which is usually accomplished by using more copper and iron, and also using
special cooling mechanisms such as oil flow and forced air.

angle = arcsin( 1 - Rf*(1-Vd/Vpk) )

This is with Rf being the ripple factor (not in percent
terms, obviously) and Vd being the sum of the diode drops
(full wave would be something like 2V) and Vpk being the
sqrt(2)*Vrms. At least, that's what the equation works out
for me on paper. (I can develop out here, if needed.)

The peak diode current happens just as the first moment of
conduction (which is neatly defined by the ripple factor, if
I understand you) and would be something like:

Ipk = 2*pi*f*C*Vpk*cos( angle )

Since the cos(arcsin(x)) is just sqrt(1-x^2), the computation
looks like:

Ipk = 2*pi*f*C*Vpk*SQRT(1-(1-Rf*(1-Vd/Vpk))^2)

There's probably some other adjustments to nail it, but that
probably gets somewhat close.

Maybe it is useful to work out these equations to get a concept of what is
going on, but I prefer a more empirical method which may involve initial
rough estimates and prototyping and bench testing, as well as LTSpice
simulation. The simulator includes the equations that determine the
performance of the circuit, and may also include the effects of losses and
heating and temperature change. But usually I just use approximations and
best guesses of final operating conditions such as temperature, and use
parameters such as internal resistance based on these figures. Then it is
time to build the circuit and do real world bench testing.

[snip]

Can you expand a little on what you were talking about,
though? Was that a half wave suggestion? I'm not sure I can
make sense of the rails, if so. If not, then I'd still
appreciate some of the calculations so that I can sure I
follow all of it.

I'm guessing that if a rail is to have a minimum of 25V on it
at the bottom of the ripple, and you are talking about 15%
regulation, the peak is going to be 29.4V -- not counting the
diode drops. Add 1V for that and it's 30.4V. Another 7% on
top would be 32.7V for the peak. RMS would be that figure
divided by sqrt(2), wouldn't it? Or 23.1Vrms or so?

So would that suggest two of the cheap 25.2Vrms transformers
and plan on rails still slightly higher?

Oh, crap. The VA rating. That's another one to consider.
Later, I guess.

I sense a lack of a real direction or intended purpose for this project. As
an academic exercise and learning experience, throwing all sorts of ideas
into the pot is worthwhile. But when it comes to the actual task of
building something useful, whether for production or a one-off hobby
project, it comes down to the three factors I offer. I can build it well, I
can build it quickly, and I can build it cheaply. Pick any TWO!

A dual 12.6 volt transformer would give a minimum (worst case with
transformer at full load and mains 7% under voltage) of 16.something
volts meaning big filter caps if you were serious at getting 10 watts.
One of the reasons to go PSU first I think. (Also I live in a tiny
jerk-water town where no one knows what a custom made transformer is let
alone where you can get one wound)

I suppose the old filament-type 12.6VAC transformers must be
common everywhere. I also see that Radio Shack (yes, I'm
holding my nose for a moment) still carries some "commodity"
type 25.2VAC CT 2.0A rated transformers for about US$10.
Their 12.6VAC CT 3.0A transformers are priced identically.

So what's considered to be generally available?

Certainly this depends on your location as well as your budget (time and/or
money) and criteria for the design. If you plan to go the cheapest monetary
route for a one-off project, look for locally available freebies in a
junkyard, flea markets, Hamfests, eBay, and www.freecycle.com. You also
must consider time and transportation or shipping expenses, which can be
high for items like transformers.

You must also balance what is readily available with what you actually need
for your project. If you have certain constraints and absolute design
criteria, you may be forced into a narrow range of what is acceptable. At
some point, you may need to modify a salvaged transformer or wind your own
(or have one custom made). There are many off-the-shelf transformers
available at reasonable cost, so it would be rare to need a custom design,
but sometimes it is the only option. You can do a lot with a MOT if you
don't mind spending the time messing with it.

And you can also get toroid transformer kits that have the primary already
wound, and you just add your own secondary. See www.toroid.com. They have
kits from 80VA ($52) to 1400VA ($110). I used four of the largest ones to
make a circuit breaker test set with an output of 2000 amperes at 2.8 volts
continuous, and the good regulation allowed it to provide pulses of over
12,000 amperes. If you find any equipment with toroid transformers, by all
means salvage them. You can also use Variacs and Powerstats and their
equivalents to make high power transformers. I have about a dozen damaged
units rated at 240 VAC at 8 amps, or 2 kVA, and I had plans to use them for
a 24 kVA test set, 4000 amps at 6 volts. Here are pictures of a 10 kVA test
set I designed for www.etiinc.com, using toroids:

http://www.smart.net/~pstech/PI2000-1-small.JPG
http://www.smart.net/~pstech/PI2000-2-small.JPG
http://www.smart.net/~pstech/PI2aux-5a.JPG

But I have digressed, and this thread has digressed from the discussion of
amplifiers to power supplies (which is related, of course), and line
powered transformers (which may not be the best choice). However, at some
point one must decide if this is to be an actual project or just an
academic discussion, and then proceed to get some parts and put something
together and plug it in. It can be done using as many "free" parts as
possible, or from the standpoint of what is the most cost-effective
overall, and in either case one must have a clear view of the end result.

Paul

 

[Jon Kirwan]

On Thu, 04 Feb 2010 05:34:52 -0800, I wrote:

On Thu, 04 Feb 2010 05:46:33 +1000, David Eather
eather@tpg.com.au> wrote:

A 30 volt CT transformer with 15% regulation and 7% mains over-voltage,
less voltage drop for the diode bride would give rails of +/- 25.

I'm not entirely familiar with the formal terms, but I figure
the 15% regulation you mentioned above must mean that the
ripple voltage goes from 100% to 85%. In general, the
equation for the angle would be something like:

angle = arcsin( 1 - Rf*(1-Vd/Vpk) )

This is with Rf being the ripple factor (not in percent
terms, obviously) and Vd being the sum of the diode drops
(full wave would be something like 2V) and Vpk being the
sqrt(2)*Vrms. At least, that's what the equation works out
for me on paper. (I can develop out here, if needed.)

The peak diode current happens just as the first moment of
conduction (which is neatly defined by the ripple factor, if
I understand you) and would be something like:

Ipk = 2*pi*f*C*Vpk*cos( angle )

Since the cos(arcsin(x)) is just sqrt(1-x^2), the computation
looks like:

Ipk = 2*pi*f*C*Vpk*SQRT(1-(1-Rf*(1-Vd/Vpk))^2)

There's probably some other adjustments to nail it, but that
probably gets somewhat close.

For a full wave bridge (I know, I think you were talking
about a half wave,

Actually, the only thing you could have been talking about is
a full wave bridge + a CT transformer to get to two rails and
ground. I knew that was the only way to get there, too, and
that's why I was staying on a bridge form of it. But I was
sadly not thinking clearly about your writing when I read it.
That's entirely my fault, of course.

but let's go with this for a moment) with
Vpk=25.2*SQRT(2) and Vd=2V (for two diodes in conduction in
the bridge) and Rf=.15 and f=60 (US-centric) and let's say a
C=2200uF, that gets something like Ipk = 15.2A.

None of this takes into account the average load current or
peak load currents. It just assumes that the ripple factor
is somehow known to be correct. I started out assuming that
the average load current would be defined entirely by the
droop (which is, of course, based upon the ripple factor
assumption) and the time from the peak of a previous cycle to
the point at which the above angle occurs and conduction
again begins. But it is complicated slightly by the fact
that the transformer supplies the entire current draw (if it
can) for a short part of the cycle _after_ the peak, as its
slope is less than the droop slope of the cap. I played with
accounting for all that but then decided that in most
practical cases the droop is hopefully not too excessive and
if not, then the angle after the peak isn't that much.. maybe
5 to 8 degrees or so... So I decided to ignore it and just
rely upon the capacitor's droop only:

Iave = C*Vpk*Rf/((pi/2+angle)/(2*pi*f))

So in the case just mentioned, I get Iave = 1.7A. I take it
that Ipk can easily be a factor of 9 or 10 greater. Also, I
note that it would probably be helpful to have nifty charts
of some kind to help pick off details like these.

Can you expand a little on what you were talking about,
though? Was that a half wave suggestion? I'm not sure I can
make sense of the rails, if so. If not, then I'd still
appreciate some of the calculations so that I can sure I
follow all of it.

I'm guessing that if a rail is to have a minimum of 25V on it
at the bottom of the ripple, and you are talking about 15%
regulation, the peak is going to be 29.4V -- not counting the
diode drops. Add 1V for that and it's 30.4V. Another 7% on
top would be 32.7V for the peak. RMS would be that figure
divided by sqrt(2), wouldn't it? Or 23.1Vrms or so?

I should have used 2V here, too. A bridge is the way to go
and that's two drops. No need to suddenly insert a half wave
thing here when massaging the numbers around. So it should
be something more like 31.4V, with 7% more being 33.8V with
margin added. And that is about 23.9Vrms, I think.

Luckily, it doesn't change the thought about 25.2Vrms xfrmrs.
(You might, of course.)

Jon

So would that suggest two of the cheap 25.2Vrms transformers
and plan on rails still slightly higher?

Oh, crap. The VA rating. That's another one to consider.
Later, I guess.

A dual 12.6 volt transformer would give a minimum (worst case with
transformer at full load and mains 7% under voltage) of 16.something
volts meaning big filter caps if you were serious at getting 10 watts.
One of the reasons to go PSU first I think. (Also I live in a tiny
jerk-water town where no one knows what a custom made transformer is let
alone where you can get one wound)

I suppose the old filament-type 12.6VAC transformers must be
common everywhere. I also see that Radio Shack (yes, I'm
holding my nose for a moment) still carries some "commodity"
type 25.2VAC CT 2.0A rated transformers for about US$10.
Their 12.6VAC CT 3.0A transformers are priced identically.

So what's considered to be generally available?

Jon

 

[Jon Kirwan]

On Thu, 4 Feb 2010 19:01:42 -0500, "Paul E. Schoen"
<paul@peschoen.com> wrote:

"Jon Kirwan" <jonk@infinitefactors.org> wrote in message
news:upmjm55fpn755sdshfljjk6r7tdk9i6ug9@4ax.com...
On Thu, 04 Feb 2010 05:46:33 +1000, David Eather
eather@tpg.com.au> wrote:

A 30 volt CT transformer with 15% regulation and 7% mains over-voltage,
less voltage drop for the diode bride would give rails of +/- 25.

I'm not entirely familiar with the formal terms, but I figure
the 15% regulation you mentioned above must mean that the
ripple voltage goes from 100% to 85%. In general, the
equation for the angle would be something like:

Transformers are specified with a percentage regulation which means the
change in voltage from no load to full load conditions. A small,
inexpensive transformer might have 15% regulation so that the 30VCT unit
would have a 15% higher output voltage with no load, or 34.5 VRMS or 48.8
P-P. Mains voltage may vary +/- 7% or 120 VAC +/- 8, or 112 to 128 VAC. At
the high end of this range the tranny puts out about 52.2 V P-P. Assuming a
FWB rectifier and the CT as reference, with 0.7 V diode drop, you get 25.4
volts peak.

Hmm. Totally new thoughts. So that's what the term
"regulation" means. It's about the transformer design? And
here I was off and away on the capacitive-filtered ripple
side. Well, that's still useful to have gone back to,
anyway.

I'm not buying the 0.7V diode drop, yet. At peak currents
near 10 times larger than average load currents, I have to
imagine more than 0.7V drop with anything silicon and not
schottky. Do they use schottky's? (Leakage comes to mind.)

Okay. So the 25V was specifying the peak, not the bottom
side. And that is unloaded, basically. Which brings up the
question of what exactly does 15% regulation _actually_ mean.
What is the definition of "full load?" Since the peak diode
currents can be quite a lot more than the average load
current from my calculations, that seems to place quite a
burden on the transformer ratings.

So could you go further here? In other words, let's say I
know that the average load current will be 1.4A, but that the
peak diode current given the bridge/capacitor design will be
15A. The transformer is a 25.2Vrms CT unit. The DC rails
are at -15 and +15, with 2200uF caps on each side to ground,
and the ripple on them is about 3.8V peak to peak (+/-1.9V
around 15V.)

What's the VA rating here? And "regulation" number are you
looking for in the transformer and how does it relate back to
VA and other terms that might be used?

If you put a capacitor on the output, it eventually charges to the peak
voltage. This is the high limit that must be considered for design. It may
not be exact, and probably will be a bit lower, because a power transformer
is usually designed to operate in partial saturation, so the output will
not increase linearly above its design rating.

Ah. Core saturation is __intended__ as part of the design? I
haven't done that one before. What guidance can you give on
that aspect?

Under load, the output will drop, caused by the effects of primary and
secondary coil resistance as well as magnetic effects. These will cause
heating over a period of time, and the coil resistance will increase,
adding to the effect until a point of equilibrium is reached based on the
ambient conditions and removal of heat via conduction, convection, and
radiation.

Now that, I understand and worry about.

Large power transformers, high quality audio transformers, and
instrumentation transformers are designed with perhaps 1% or 2% regulation,
which is usually accomplished by using more copper and iron, and also using
special cooling mechanisms such as oil flow and forced air.

Okay.

angle = arcsin( 1 - Rf*(1-Vd/Vpk) )

This is with Rf being the ripple factor (not in percent
terms, obviously) and Vd being the sum of the diode drops
(full wave would be something like 2V) and Vpk being the
sqrt(2)*Vrms. At least, that's what the equation works out
for me on paper. (I can develop out here, if needed.)

The peak diode current happens just as the first moment of
conduction (which is neatly defined by the ripple factor, if
I understand you) and would be something like:

Ipk = 2*pi*f*C*Vpk*cos( angle )

Since the cos(arcsin(x)) is just sqrt(1-x^2), the computation
looks like:

Ipk = 2*pi*f*C*Vpk*SQRT(1-(1-Rf*(1-Vd/Vpk))^2)

There's probably some other adjustments to nail it, but that
probably gets somewhat close.

Maybe it is useful to work out these equations to get a concept of what is
going on, but I prefer a more empirical method which may involve initial
rough estimates and prototyping and bench testing, as well as LTSpice
simulation.

I reverse this. I like understanding the _theory_ and don't
care at all about practice until _after_ I've mastered the
theoretical aspects that bear more on the problems. I _then_
use LTspice _after_ being able to work things on paper, just
to check and verify that I got it. The reason is, if I'm
missing something important it will then show up and that
will kick me to go back and find additional theory to cover
the gap in my paper knowledge. That's how I learn. It's the
only way I really feel that I understand something. (I think
I talked a little about that here.)

The simulator includes the equations that determine the
performance of the circuit, and may also include the effects of losses and
heating and temperature change. But usually I just use approximations and
best guesses of final operating conditions such as temperature, and use
parameters such as internal resistance based on these figures. Then it is
time to build the circuit and do real world bench testing.

I think there is always time to go build. And when I do
that, I will take measurements and make adjustments to get
where I want to be and I won't be sweating the theory so much
at that point. However, before I get there I like to make
sure I've mastered the relevant theories.

Let me put this in an entirely different context that may
shed some light on "how I think" and "why I think that way."

The reality of modern US surgery is that an anesthesiologist
uses well-worn practice with well-surveyed and well-studied
drugs and tools. They work. And as a general matter, they
work most of the time without the anesthesiologist having to
remember anything about chemistry or metabolites or liver
pathways or the kidney micropipette filtering system. They
don't care about memorizing any of that, or frankly, even
knowing much how it works.

I make it a matter of regular practice to check off the "yes"
box every time any of my family gets the surgery forms where
there always present the question, "Do you want to meet with
the anesthesiologist?" The very first question I ask is,
"What are you using and what are the liver pathways and
resulting metabolites for it?" I have yet to have a single
one of them be able to answer the question. Not once. I
have had one or two tell me that "Well, we studied all that
in school but I don't remember any of it." At least a frank
admission there.

So why do I care? I completely understand that in almost
every case on the operating table there will be no problems
and that the well-worn paths in anesthesia work on most
people most of the time. However, there is a reason. What
if something unusual takes place. A unique reaction, for
example. Something _outside_ the usual experience. What
then? They would have, let's say, minutes to make a
decision. No time to go to books.

I want someone there who knows the chemistry, knows what _is_
known about the pathways. How much of the primary pathway is
used in proportion to the other pathways? What are the
products in the other pathways? What are their effects
should they exceed some limit? How does that present or
manifest itself?

These are the kinds of things that might bring to bear an
answer -- something needed to mitigate a disaster in the
making when there is no time to hit a book but where if they
did understand the theories well and knew the pathways and
the effects of excessive amounts of the unusual pathway
metabolites they might know exactly what to do when it would
mean the difference of life and death.

There are MANY people who die in these circumstances that are
chocked up to "Oh, well. It happens rarely."

If I were an anesthesiologist, Paul, I'd know this stuff
cold. And I'd keep up on the current knowledge, too. And
more. Because that's the way I am.

Yes, they are practical people and they do a very
satisfactory professional job every day of their lives. But
quite frankly I don't think that's good enough.

Theory provides _all_ meaning. And it's the way I think
about things. It's how I function. Yes, others will be very
satisfied with "practical" results. I'm not. I need more.

Can you expand a little on what you were talking about,
though? Was that a half wave suggestion? I'm not sure I can
make sense of the rails, if so. If not, then I'd still
appreciate some of the calculations so that I can sure I
follow all of it.

I'm guessing that if a rail is to have a minimum of 25V on it
at the bottom of the ripple, and you are talking about 15%
regulation, the peak is going to be 29.4V -- not counting the
diode drops. Add 1V for that and it's 30.4V. Another 7% on
top would be 32.7V for the peak. RMS would be that figure
divided by sqrt(2), wouldn't it? Or 23.1Vrms or so?

So would that suggest two of the cheap 25.2Vrms transformers
and plan on rails still slightly higher?

Oh, crap. The VA rating. That's another one to consider.
Later, I guess.

I sense a lack of a real direction or intended purpose for this project.

It's for education. I think I stated that at the outset. I
sure hope I did. However, I _do_ intend on producing a
practical result. Not because I need one. But because I
need to make sure that my mental models work, in real
practice. It's like some theorist saying that if you pass
electrons by a certain kind of magnetic field, they will
separate according to a certain observational spin. Great.
But until you build and test the idea, you really don't know.
So you build and test. I intend to build and test an
amplifier, not because I need one badly, but because I want
to see how all that theory works in practical building
circumstances. It may highlight yet something new that I
hadn't considered and will point me to still more theory to
gain a hold upon.

Where this knowledge will wind up "doing something" is
unknown at this stage. It might get used in ways two years
from now that I have no way to predict, today. Or ten years.
I hope to live long enough to see some utility, though it may
not be with amplifiers.

However, I had posted a different question a while back about
my autistic daughter's abuse of volume controls around the
house and ultimately I hope to use this knowledge in
designing a custom system for her that does include some
features probably few others will care about. So I foresee
something in the next year, to be honest.

But the main point is learning, right now. I need to grasp
this stuff from start to end to some _reasonable_ level.

As
an academic exercise and learning experience, throwing all sorts of ideas
into the pot is worthwhile.

That's me.

But when it comes to the actual task of
building something useful, whether for production or a one-off hobby
project, it comes down to the three factors I offer. I can build it well, I
can build it quickly, and I can build it cheaply. Pick any TWO!

Hehe. I want to _learn_ to design to specified criteria,
have a comprehensive view of the theoretical concepts
involved, and that means I need to only pick the first one.
The 'quickly' is unimportant -- one to two years is good
enough. The 'cheaply' is equally unimportant. If it costs
me 10 times as much in terms of parts and time as it would
just buying something commercial, buying a commercial
solution will teach me exactly zero about what I need to
learn to design what my daughter needs. And there is NOTHING
on the market to get there, either. No one else has my
problem. Or few do.

This is a "give a person a fish and they eat for a day, teach
a person to fish and they eat for the rest of their lives"
thing. It won't just apply to the next solution for my
daughter. It will help me in other ways I poorly understand
right now.

A dual 12.6 volt transformer would give a minimum (worst case with
transformer at full load and mains 7% under voltage) of 16.something
volts meaning big filter caps if you were serious at getting 10 watts.
One of the reasons to go PSU first I think. (Also I live in a tiny
jerk-water town where no one knows what a custom made transformer is let
alone where you can get one wound)

I suppose the old filament-type 12.6VAC transformers must be
common everywhere. I also see that Radio Shack (yes, I'm
holding my nose for a moment) still carries some "commodity"
type 25.2VAC CT 2.0A rated transformers for about US$10.
Their 12.6VAC CT 3.0A transformers are priced identically.

So what's considered to be generally available?

Certainly this depends on your location as well as your budget (time and/or
money) and criteria for the design. If you plan to go the cheapest monetary
route for a one-off project, look for locally available freebies in a
junkyard, flea markets, Hamfests, eBay, and www.freecycle.com. You also
must consider time and transportation or shipping expenses, which can be
high for items like transformers.

I was about to write, earlier, that I already have a large
supply of scavenged transformers. I will have _no_ problem
finding a suitable one somewhere in the pile. The question
was brought up by David. So I asked, that's all.

You must also balance what is readily available with what you actually need
for your project. If you have certain constraints and absolute design
criteria, you may be forced into a narrow range of what is acceptable. At
some point, you may need to modify a salvaged transformer or wind your own
(or have one custom made). There are many off-the-shelf transformers
available at reasonable cost, so it would be rare to need a custom design,
but sometimes it is the only option. You can do a lot with a MOT if you
don't mind spending the time messing with it.

For now, I'm just planning to use what I can lay hands on...
when the time comes. However, I don't mind at all any
discussion about practical choices were someone to buy new
parts. That teaches me about others and their concerns and
helps me to help others, too.

In short, this topic is made even better when it isn't just
about me and my interests and my focus. I like it very much
when others chip in about other thoughts, other places and
times, and where ever that may take it. However, I am still
clear about what part of it makes the most difference for me
-- the learning part about everything from power supply
design, input stage design, class A, class B, class AB
considerations, output stages and drivers, VAS, splitters,
current mirrors, current sources, etc. It's all good to me.

That part of this discussion that went off on the direction
of ICs was also fine. I took note and figure on getting back
to thinking about that too, someday later on.

And you can also get toroid transformer kits that have the primary already
wound, and you just add your own secondary. See www.toroid.com. They have
kits from 80VA ($52) to 1400VA ($110). I used four of the largest ones to
make a circuit breaker test set with an output of 2000 amperes at 2.8 volts
continuous, and the good regulation allowed it to provide pulses of over
12,000 amperes. If you find any equipment with toroid transformers, by all
means salvage them. You can also use Variacs and Powerstats and their
equivalents to make high power transformers. I have about a dozen damaged
units rated at 240 VAC at 8 amps, or 2 kVA, and I had plans to use them for
a 24 kVA test set, 4000 amps at 6 volts. Here are pictures of a 10 kVA test
set I designed for www.etiinc.com, using toroids:

http://www.smart.net/~pstech/PI2000-1-small.JPG
http://www.smart.net/~pstech/PI2000-2-small.JPG
http://www.smart.net/~pstech/PI2aux-5a.JPG

But I have digressed, and this thread has digressed from the discussion of
amplifiers to power supplies (which is related, of course), and line
powered transformers (which may not be the best choice). However, at some
point one must decide if this is to be an actual project or just an
academic discussion, and then proceed to get some parts and put something
together and plug it in. It can be done using as many "free" parts as
possible, or from the standpoint of what is the most cost-effective
overall, and in either case one must have a clear view of the end result.

The digressions are great! I am NOT in a rush to build,
though. I'm wanting to engage the math and learn what can be
achieved by deducing from parsimonous theory. Then test a
few things on the bench, ask questions, learn some more. Etc.
So theory _and_ practical approaches are important. Not one,
or the other, but both!!

Pendulum motion is well understood. One might either have a
practical knowledge about it and some tables and just go with
that. Probably, lots of folks making pendulum clocks stop
there and go no further and are none the worse for that. It
is similarly very easy to develop the infinite series that
describes it (or use the sqrt(L/g) proportionality as a first
order approximation or for small starting angles) from the
simple differentials involved and to take an entirely
theoretical approach, as well.

But I'm interested in more than that. Theory by itself lacks
reality. Reality by itself lacks meaning sans theory. The
two go together like hand in glove, though. Building even
the most simple ones using a peg-in-hole method leads to the
discovery of still more interesting effects, if you know some
theory. For example, the rocking of the pin itself in the
larger hole has a measurable impact of perhaps as much as 2
or 3 percent. It's useful to know that and understand it.
Once that mechanism is itself understood, one can then dig
even deeper to find more subtle (and possibly useful) effects
to continue improvements. A practitioner lacking even the
basic theory might accidentally happen upon some idea, of
course. And a theoretician lacking practical reality to
interfere might accidentally imagine some realistic effect to
pursue, too. But it really takes a marriage of both to make
quick work of progress forward, I think.

Since theory is primary, I like to pursue that part of it
earlier and move to experience once I have the mental tools
required to make sense of the data that results. Without
theory, data is pure noise. Without the theory of a sphere,
even the gentle curvature at the horizon "seen" my a mountain
climber is just so much useless noise to them. But _with_
that theory, the data _means_ much.

Jon

 

[Jon Kirwan]

On Thu, 04 Feb 2010 17:16:59 -0800, I wrote:

snip
Since theory is primary, I like to pursue that part of it
earlier and move to experience once I have the mental tools
required to make sense of the data that results.
snip

Okay. On second thought... enough theory. I think it's time
for practice. I already have triple output power supplies,
but using them wouldn't be true to the actual amplifier
situation. And any testing of distortions needs to cope with
that reality.

So I'm moving forward on the power supply rails. I need to
scarf around and see what I have available. I'll post what I
find, the resulting design and thinking, photos perhaps, and
the results of testing with static loads. Once that is done,
I'd like some advice about the next step, though. But until
then, I'll just focus on getting that part put to bed. That
much I can do right now.

I've decided that your kick in the butt, Paul, was what I
needed. I have enough in mind to move out of the thinking
stage and into trying some different alternatives. I'll get
going.

Thanks,
Jon

 

[Paul E. Schoen]

"Jon Kirwan" <jonk@infinitefactors.org> wrote in message
news:iqomm597nf576piiskbpq3m14uuiheont3@4ax.com...

Hmm. Totally new thoughts. So that's what the term
"regulation" means. It's about the transformer design? And
here I was off and away on the capacitive-filtered ripple
side. Well, that's still useful to have gone back to,
anyway.

I'm not buying the 0.7V diode drop, yet. At peak currents
near 10 times larger than average load currents, I have to
imagine more than 0.7V drop with anything silicon and not
schottky. Do they use schottky's? (Leakage comes to mind.)

Silicon diodes are the norm except for high power, high efficiency, high
frequency, and low voltage. But they do have forward drops of 0.7 to 0.6
volts at normal operating temperatures, and when drawing minimal current,
as is the case at the waveform peak under no load conditions. Even with a
capacitor, the diode current drops to near zero at the voltage peak. A
different result is expected if there is inductance, of course.

There is a separate regulation spec for the DC output. It is typically much
worse than the regulation of the transformer, as the capacitors quickly
discharge between peaks and can be charged up only as quickly as the
transformer and diodes allow during the conduction cycle. So we use big
capacitors and linear regulators, or resort to a switching supply.

But if you are lucky enough to have three phase power, you can design a DC
supply with no capacitors and get something like 6% regulation (and
ripple). This is SOP for really high power DC, like 10kVA.

Okay. So the 25V was specifying the peak, not the bottom
side. And that is unloaded, basically. Which brings up the
question of what exactly does 15% regulation _actually_ mean.
What is the definition of "full load?" Since the peak diode
currents can be quite a lot more than the average load
current from my calculations, that seems to place quite a
burden on the transformer ratings.

Transformers are rated at RMS current, which is pretty much all that
matters for heating effect, and it is mostly related to the resistance of
the copper and the allowable rise in temperature in the core. Efficiency
aside, what matters is the temperature the insulation can withstand before
deteriorating, and usually that is at least 130C, or 100C above ambient.
The smaller the tranny, the better it sheds heat (surface area/volume), so
regulation and efficiency of smaller ones tend to be poorer.

Full load is just the maximum RMS current at which the transformer is
rated. This may be further complicated by duty cycle ratings, which can be
continuous or intermittent. Generally intermittent duty is 50% duty cycle,
with ON times not greater than 30 minutes, at least for larger transformers
with more thermal mass. At 50% duty cycle the output rating is 1.4 times
the true continuous rating. And then the allowable duty cycle is the
inverse of the square of the overload. For the circuit breaker test sets I
design, we specify output up to 10x the continuous rating, at which the
duty cycle is only 1%. But the ON time is limited to about 100 mSec, which
is more than enough to trip a circuit breaker instantaneously, and then you
should wait 10 seconds before doing it again.

I designed a "Programmable Overload Device", or POD, which takes into
account the current and the time, as well as the actual temperature using a
thermistor, to enforce reasonable duty cycles. Fuses, circuit breakers, and
Motor Overloads do a similar function, but don't fully take into account
all the factors. The intelligence for this is buried in the PIC code, and
is rather involved and yet imperfect. If I could accurately model the
heating and cooling effects of current in a transformer, it would be ideal.
Now that's where theory can really help.

So could you go further here? In other words, let's say I
know that the average load current will be 1.4A, but that the
peak diode current given the bridge/capacitor design will be
15A. The transformer is a 25.2Vrms CT unit. The DC rails
are at -15 and +15, with 2200uF caps on each side to ground,
and the ripple on them is about 3.8V peak to peak (+/-1.9V
around 15V.)

What's the VA rating here? And "regulation" number are you
looking for in the transformer and how does it relate back to
VA and other terms that might be used?

It's really easier (and perhaps even more appropriate) to use a tool such
as LTSpice for this purpose. You could look at all the variables over time,
quantized to steps small enough to minimize error, and finally arrive at a
steady state solution where you may be able to describe such complex
entities as RMS current with an equation, but all you will have done is
spend a lot of time doing what LTSpice does so well and so quickly. So I
cobbed together a simple power supply simulation, which in this case models
part of a power supply that I have been using on my Ortmasters, with a
Signal 241-6-16 transformer. The ASCII file is at the end of this post.

I'm using a voltage doubler circuit on each leg of the 16VCT transformer,
as I need to get at least 17 VDC for 15VDC linear regulators for the analog
portion of the circuit. I figure no more than 20 mA. So for simulation
puposes I use a 1k resistor as the load. The transformer is 32 VA, or 2A at
16V, and I estimate 15% regulation which is a 2.4 V drop at 16V or open
circuit 18.4 VRMS. I'm using a voltage source with 26 volts peak and 1.1
ohms internal resistance. The capacitors are 220uF, and MURS120 diodes. As
a result, I get 22.35 VDC outputs, and the transformer current is 104 mA
RMS, with peaks of about 360 mA.

Just for fun, I changed the output loads to 10 ohms, and I found that the
current is only 345 mA RMS, and the transformer current is 611 mA RMS, with
peaks of about 1 amp. The capacitively coupled design is inherently
current-limited, which can be a good thing.

If you put a capacitor on the output, it eventually charges to the peak
voltage. This is the high limit that must be considered for design. It
may
not be exact, and probably will be a bit lower, because a power
transformer
is usually designed to operate in partial saturation, so the output will
not increase linearly above its design rating.

Ah. Core saturation is __intended__ as part of the design? I
haven't done that one before. What guidance can you give on
that aspect?

Maximum use of the iron occurs near the maximum flux density. It results in
increased current which actually occurs at 90 degrees to the applied
voltage, so the distortion is not in the form of a flattening of the
voltage waveform but rather like crossover distortion. But it does result
in a somewhat non-linear effect, as it interacts with the resistance of the
windings. See the following for more information:
http://openbookproject.net/electricCircuits/AC/AC_9.html

and more about regulation:
http://www.allaboutcircuits.com/vol_2/chpt_9/6.html

It is most pronounced in ferroresonant transformers:
http://www.ustpower.com/Support/Voltage_Regulator_Comparison/Ferroresonant_Transformer_CVT/Constant_Voltage_Transformer_Operation.aspx

Under load, the output will drop, caused by the effects of primary and
secondary coil resistance as well as magnetic effects. These will cause
heating over a period of time, and the coil resistance will increase,
adding to the effect until a point of equilibrium is reached based on the
ambient conditions and removal of heat via conduction, convection, and
radiation.

Now that, I understand and worry about.

That's why most designs are made with a generous safety factor so you do
not need to worry about these effects. They can be predicted approximately
and that is good enough.

Hehe. I want to _learn_ to design to specified criteria,
have a comprehensive view of the theoretical concepts
involved, and that means I need to only pick the first one.
The 'quickly' is unimportant -- one to two years is good
enough. The 'cheaply' is equally unimportant. If it costs
me 10 times as much in terms of parts and time as it would
just buying something commercial, buying a commercial
solution will teach me exactly zero about what I need to
learn to design what my daughter needs. And there is NOTHING
on the market to get there, either. No one else has my
problem. Or few do.

It might be worthwhile to discuss those details here to dig up some ideas.

This is a "give a person a fish and they eat for a day, teach
a person to fish and they eat for the rest of their lives"
thing.

I've heard it said that, "teach a man to fish, and he'll spend all day in a
boat drinking beer!"

The digressions are great! I am NOT in a rush to build,
though. I'm wanting to engage the math and learn what can be
achieved by deducing from parsimonous theory. Then test a
few things on the bench, ask questions, learn some more. Etc.
So theory _and_ practical approaches are important. Not one,
or the other, but both!!

Pendulum motion is well understood. One might either have a
practical knowledge about it and some tables and just go with
that. Probably, lots of folks making pendulum clocks stop
there and go no further and are none the worse for that. It
is similarly very easy to develop the infinite series that
describes it (or use the sqrt(L/g) proportionality as a first
order approximation or for small starting angles) from the
simple differentials involved and to take an entirely
theoretical approach, as well.

But I'm interested in more than that. Theory by itself lacks
reality. Reality by itself lacks meaning sans theory. The
two go together like hand in glove, though. Building even
the most simple ones using a peg-in-hole method leads to the
discovery of still more interesting effects, if you know some
theory. For example, the rocking of the pin itself in the
larger hole has a measurable impact of perhaps as much as 2
or 3 percent. It's useful to know that and understand it.
Once that mechanism is itself understood, one can then dig
even deeper to find more subtle (and possibly useful) effects
to continue improvements. A practitioner lacking even the
basic theory might accidentally happen upon some idea, of
course. And a theoretician lacking practical reality to
interfere might accidentally imagine some realistic effect to
pursue, too. But it really takes a marriage of both to make
quick work of progress forward, I think.

Since theory is primary, I like to pursue that part of it
earlier and move to experience once I have the mental tools
required to make sense of the data that results. Without
theory, data is pure noise. Without the theory of a sphere,
even the gentle curvature at the horizon "seen" my a mountain
climber is just so much useless noise to them. But _with_
that theory, the data _means_ much.

I think I had problems in the EE program at Johns Hopkins because it was
too theoretical for my mindset, and I had fundamental problems with
advanced calculus. I aced the lab courses and helped others because I had
already designed and built many circuits. But, looking back, I see where
having a stronger grasp of theory would have helped. I still design
circuits with a highly empirical approach, using rule of thumb and
experience to choose components. Now that SPICE is freely available I find
it fascinating to try different values and placements and configurations
"just to see what happens". And I learn by looking at the time domain
simulation plots and determining what may have caused certain glitches or
oscillations that I did not foresee.

My talents are more in the realm of imagination and thinking outside the
box. And sometimes it has gotten me into trouble. But I have also sometimes
been able to make a lot of progress in a short period of time. I think some
aspects of design are more of an art than a science, and I look for a sort
of elegance in the finished design of a circuit, even in the placement of
components on the schematic, and also in their placement on a PCB.

Paul

------------------------------------------------------------------

Version 4
SHEET 1 880 680
WIRE 48 112 -96 112
WIRE 48 144 48 112
WIRE 112 144 48 144
WIRE 224 144 176 144
WIRE 288 144 224 144
WIRE 384 144 352 144
WIRE 512 144 384 144
WIRE 528 144 512 144
WIRE 528 160 528 144
WIRE 224 176 224 144
WIRE 384 176 384 144
WIRE -96 208 -96 112
WIRE 48 256 48 224
WIRE 224 256 224 240
WIRE 224 256 48 256
WIRE 256 256 224 256
WIRE 384 256 384 240
WIRE 384 256 256 256
WIRE 416 256 384 256
WIRE 528 256 528 240
WIRE 528 256 416 256
WIRE 560 256 528 256
WIRE 608 256 560 256
WIRE 48 288 48 256
WIRE 608 288 608 256
WIRE 416 304 416 256
WIRE 560 336 560 256
WIRE 256 352 256 256
WIRE -96 400 -96 288
WIRE 48 400 48 368
WIRE 48 400 -96 400
WIRE 48 480 48 400
WIRE 96 480 48 480
WIRE 256 480 256 416
WIRE 256 480 160 480
WIRE 304 480 256 480
WIRE 416 480 416 368
WIRE 416 480 368 480
WIRE 528 480 416 480
WIRE 560 480 560 416
WIRE 560 480 528 480
FLAG 608 288 0
FLAG 512 144 V+
FLAG 528 480 V-
SYMBOL ind2 64 240 R180
WINDOW 0 36 80 Left 0
WINDOW 3 36 40 Left 0
SYMATTR InstName L1
SYMATTR Value 1
SYMATTR Type ind
SYMATTR SpiceLine Rser=.01
SYMBOL ind2 64 384 R180
WINDOW 0 36 80 Left 0
WINDOW 3 36 40 Left 0
SYMATTR InstName L2
SYMATTR Value 1
SYMATTR Type ind
SYMATTR SpiceLine Rser=.01
SYMBOL polcap 176 128 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName C1
SYMATTR Value 220ľ
SYMATTR Description Capacitor
SYMATTR Type cap
SYMATTR SpiceLine V=63 Irms=2.51 Rser=0.025 Lser=0
SYMBOL diode 240 240 R180
WINDOW 0 24 72 Left 0
WINDOW 3 24 0 Left 0
SYMATTR InstName D1
SYMATTR Value MURS120
SYMBOL diode 288 160 R270
WINDOW 0 32 32 VTop 0
WINDOW 3 0 32 VBottom 0
SYMATTR InstName D2
SYMATTR Value MURS120
SYMBOL polcap 368 176 R0
SYMATTR InstName C2
SYMATTR Value 220ľ
SYMATTR Description Capacitor
SYMATTR Type cap
SYMATTR SpiceLine V=63 Irms=2.51 Rser=0.025 Lser=0
SYMBOL voltage -96 192 R0
WINDOW 3 -8 233 Left 0
WINDOW 123 0 0 Left 0
WINDOW 39 -7 260 Left 0
SYMATTR Value SINE(0 26 60 0 0 0 100)
SYMATTR SpiceLine Rser=1.1
SYMATTR InstName V1
SYMBOL res 512 144 R0
SYMATTR InstName R1
SYMATTR Value 1k
SYMBOL polcap 160 464 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName C3
SYMATTR Value 220ľ
SYMATTR Description Capacitor
SYMATTR Type cap
SYMATTR SpiceLine V=63 Irms=2.51 Rser=0.025 Lser=0
SYMBOL polcap 400 304 R0
SYMATTR InstName C4
SYMATTR Value 220ľ
SYMATTR Description Capacitor
SYMATTR Type cap
SYMATTR SpiceLine V=63 Irms=2.51 Rser=0.025 Lser=0
SYMBOL diode 368 496 M270
WINDOW 0 32 32 VTop 0
WINDOW 3 0 32 VBottom 0
SYMATTR InstName D3
SYMATTR Value MURS120
SYMBOL diode 240 416 M180
WINDOW 0 24 72 Left 0
WINDOW 3 24 0 Left 0
SYMATTR InstName D4
SYMATTR Value MURS120
SYMBOL res 544 320 R0
SYMATTR InstName R2
SYMATTR Value 1k
TEXT -104 480 Left 0 !K1 L1 L2 1
TEXT -104 512 Left 0 !.tran 1

 

[pimpom]

Jon Kirwan wrote:

On Thu, 04 Feb 2010 17:16:59 -0800, I wrote:

snip
Since theory is primary, I like to pursue that part of it
earlier and move to experience once I have the mental tools
required to make sense of the data that results.
snip

Okay. On second thought... enough theory. I think it's time
for practice. I already have triple output power supplies,
but using them wouldn't be true to the actual amplifier
situation. And any testing of distortions needs to cope with
that reality.

So I'm moving forward on the power supply rails. I need to
scarf around and see what I have available. I'll post what I
find, the resulting design and thinking, photos perhaps, and
the results of testing with static loads. Once that is done,
I'd like some advice about the next step, though. But until
then, I'll just focus on getting that part put to bed. That
much I can do right now.

I've decided that your kick in the butt, Paul, was what I
needed. I have enough in mind to move out of the thinking
stage and into trying some different alternatives. I'll get
going.

Thanks,
Jon

There's this saying "Practice without theory is blind and theory
without practice is lame". You've made it clear that you want to
thoroughly understand the hows and whys of amplifier design from
mathematical models. I have no quarrel with that approach and I
also use it myself within the limits of my own capability - *up
to a point*. But there comes a point at which striving for
absolute precision solely from theory results in diminishing
returns.

Take the case of the pendulum you brought up earlier. The basic
theory is well established, but to predict the behaviour of a
practical pendulum with 100% precision will require taking into
account the effects of so many factors that it may well be
impossible. E.g., the aerodynamics of the pendulum's shape
including minute irreguarities on its surface, the exact strength
and orientation of the earth's magnetic field at the location and
its effect on traces of magnetic materials in the alloy, friction
with suspended particles in the air in addition to the air
itself, friction at the point of suspension and elasticity of the
suspension, etc., etc. Even if all these influencing factors are
included in the equation, the physical values to be entered can
never be measured with 100% accuracy.

Take the case of the forward drop of the diode in the power
supply that you've been discussing with Paul. This what I did
before personal computers and simulation progs became widely
available: I drew a curve of the diode's V-I characteristics on
graph paper up to the expected peak current. Then I drew a
straight line, approximately following the dynamic curve, from
the peak point down to the voltage axis. I took that voltage as a
constant forward drop and the slope of the line as a constant
series resistor. I then added that resistance to other source
resistances like the transformer winding resistance and either
use it to calculate the rectified and filtered voltage or, more
often, to determine it from a graph such as that in RDH. It also
comes in useful for finding the peak and rms currents. I don't
know if anyone else uses that method or how well it agrees with
theory, but it agrees pretty well with practical measurements.

I don't do this every time I design a power supply. I just make a
mental estimate based partly on theory and partly on past
experience. In short, there's a point at which it makes more
sense to make informed assumptions and approximations even before
doing physical construction.