OK, i decided to have a little fun messing with computer cores from
the IBM 1620 etc.
Had 3 sizes to look at:
Large: 0.135 OD, ~0.072 ID, 0.071 thick (used 10T primary, 1T
secondary) [possibly could do 50T]
Medium: 0.070 OD, ~0.045 ID, 0.024 thick (used 5T) [possibly could do 20T]
Small: 0.050 OD, 0.029 ID, 0.011 thick (used 3T) [possibly could do
5T, guess absolute max 7T]
Used #36 wire as that was all i had in small (thermaleze) wire.

Started with the large core, HP3312 function generator set to sine
drive 10V pk-pk driving in series 50 ohm resistor then primary of core
to ground; Tek P6019 probe for monitoring current drive and standard
probes to monitor voltage at top of primary (and secondary in this case).
Gee aren't transformers nice?
Secondary was essentially one-tenth the amplitude of the primary at
any reasonable frequency.
So i dumped the use of a secondary for monitoring amplitude/waveshape.

Basically, "BH" loop not seen; only a nice ellipse - BUT spikes on
the end: right side positive and left side negative.
"Interesting" part was the secondary showed opposite polarity on the
spikes.
Spikes increased in amplitude with drive and/or frequency.

So, i switched from XY mode to time domain; see Large.gif attached;
65Hz drive 10mV/div vertically.
Notice the spikes are a bit after the cross-over.

Basically the same thing seen with the medium-sized core; 600Hz
drive; notice spikes a lot smaller.

Did not bother taking a pic of what i saw when trying the small core.
Spikes were very dinky, hard to find/see but consistent in
positioning and amplitude at any reasonable frequency.

So..nothing like a BH loop and spikes (??) instead, consistent in
positioning on drive waveform.

Any hints, comments or suggestions?

Looks quite good, actually. Like Fred says, integrate, then you will have
it right.

Try amorphous or nanocrystalline cores, too. They switch quite quickly.
I've built a toy circuit that generated enough harmonics to cause itself to
ring a bit; the risetime was under 1us.

Physics guys use the stuff to generate fast (1us) current pulses for
particle accelerators and junk. They drive them with 3kA+ IGBTs, which
don't even switch as fast (~1us). Lots of power density.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

"Robert Baer" wrote in message
net...
OK, i decided to have a little fun messing with computer cores from
the IBM 1620 etc.
Had 3 sizes to look at:
Large: 0.135 OD, ~0.072 ID, 0.071 thick (used 10T primary, 1T
secondary) [possibly could do 50T]
Medium: 0.070 OD, ~0.045 ID, 0.024 thick (used 5T) [possibly could do 20T]
Small: 0.050 OD, 0.029 ID, 0.011 thick (used 3T) [possibly could do
5T, guess absolute max 7T]
Used #36 wire as that was all i had in small (thermaleze) wire.

Started with the large core, HP3312 function generator set to sine
drive 10V pk-pk driving in series 50 ohm resistor then primary of core
to ground; Tek P6019 probe for monitoring current drive and standard
probes to monitor voltage at top of primary (and secondary in this case).
Gee aren't transformers nice?
Secondary was essentially one-tenth the amplitude of the primary at
any reasonable frequency.
So i dumped the use of a secondary for monitoring amplitude/waveshape.

Basically, "BH" loop not seen; only a nice ellipse - BUT spikes on
the end: right side positive and left side negative.
"Interesting" part was the secondary showed opposite polarity on the
spikes.
Spikes increased in amplitude with drive and/or frequency.

So, i switched from XY mode to time domain; see Large.gif attached;
65Hz drive 10mV/div vertically.
Notice the spikes are a bit after the cross-over.

Basically the same thing seen with the medium-sized core; 600Hz
drive; notice spikes a lot smaller.

Did not bother taking a pic of what i saw when trying the small core.
Spikes were very dinky, hard to find/see but consistent in
positioning and amplitude at any reasonable frequency.

So..nothing like a BH loop and spikes (??) instead, consistent in
positioning on drive waveform.

Any hints, comments or suggestions?

Fred Abse wrote:
On Sat, 09 Jun 2012 20:37:02 -0800, Robert Baer wrote:

Basically, "BH" loop not seen; only a nice ellipse

You won't see a B-H loop that way.

On the X axis, you have current in the winding, that's fine. proportional
to H.

The Y axis is where it goes wrong:

Now, V = -L di/dt, or V proportional to d(phi)/dt

You want B (that is to say phi) on the Y axis.

You're seeing something proportional to dB/dt

Hence you need to integrate the voltage applied to the Y input.

Similar to how you'd use a Rogowski coil.

You're seeing what I'd expect to see with your test setup.

I do not think so; what is that pulse?
WRT percentage of time of sine drive, the pulse is in same position
and same width as seen from 0.5Hz to 500Hz; pops up at 5% of half-sine
time, jumps back down at 12% of half-sine time.

So, you might say that as the current drive increases, the core
eventually saturates and thereby the inductance goes toward zero, making
the voltage rise rapidly - that can explain the rise and top of the pulse.
BUT..why does it drop rapidly back after a pre-determined _increase_
of drive?
This is independent of frequency, so inductance, impedance are
meaningless arguments especially in the light of the frequency range
used for investigation.

Found a problem with my layout and resolved it.
Primary and secondary are identical - including the pulse.

On Sun, 10 Jun 2012 11:25:54 -0500, "Tim Williams"
wrote:

Looks quite good, actually. Like Fred says, integrate, then you will have
it right.

Try amorphous or nanocrystalline cores, too. They switch quite quickly.
I've built a toy circuit that generated enough harmonics to cause itself to
ring a bit; the risetime was under 1us.

Physics guys use the stuff to generate fast (1us) current pulses for
particle accelerators and junk. They drive them with 3kA+ IGBTs, which
don't even switch as fast (~1us). Lots of power density.

Tim

Nonlinear transmission lines are cool. You can make them with
saturating inductors, ceramic capacitors, or varicap diodes, even
ordinary power diodes. You can get the risetime to go down as a step
propagates down the line.


--

John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

"John Larkin" wrote in message
...
Nonlinear transmission lines are cool. You can make them with
saturating inductors, ceramic capacitors, or varicap diodes, even
ordinary power diodes. You can get the risetime to go down as a step
propagates down the line.

I've been tempted to try that before -- a necklace of ferrite beads might
run you only a couple bucks per foot, and with the velocity factor so low,
that's enough to get something snappy with jellybean transistors (a few
nanoseconds' input edge). Downside is the impedance, which must be in the
kohms range (small signal). Hard to get much current into it like that.
Supposedly, ferrite has a high e_r as well, so the impedance probably isn't
*that* bad, but it's hard to ensure a consistent airgap (or lack thereof)
with loose beads.

I expect the pulse width is limited as much by geometry (ferrite beads being
a coax structure, periodically disturbed by the imperfect faces between
beads and imperfect spacing to the conductors) as by material dispersion
(losses, etc).

The not-even-ludicrous speed[1] generators do it with schottky junctions in
InP (like the LeCroy 200GHz scope) IIRC. With harmonics near the THz, there
isn't any physical length to spare before the signal dissipates, anyway.

Tim

[1]
"Aah! What the hell was that?!"
"Spaceball 1!"
"They've gone to plaid!"

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

Tim Williams wrote:
Looks quite good, actually. Like Fred says, integrate, then you will
have it right.

Try amorphous or nanocrystalline cores, too. They switch quite quickly.
I've built a toy circuit that generated enough harmonics to cause itself
to ring a bit; the risetime was under 1us.

Physics guys use the stuff to generate fast (1us) current pulses for
particle accelerators and junk. They drive them with 3kA+ IGBTs, which
don't even switch as fast (~1us). Lots of power density.

Tim

Where can i buy some?
Are there various sizes available?

John Larkin wrote:
On Sun, 10 Jun 2012 11:25:54 -0500, "Tim Williams"
wrote:

Looks quite good, actually. Like Fred says, integrate, then you will have
it right.

Try amorphous or nanocrystalline cores, too. They switch quite quickly.
I've built a toy circuit that generated enough harmonics to cause itself to
ring a bit; the risetime was under 1us.

Physics guys use the stuff to generate fast (1us) current pulses for
particle accelerators and junk. They drive them with 3kA+ IGBTs, which
don't even switch as fast (~1us). Lots of power density.

Tim

Nonlinear transmission lines are cool. You can make them with
saturating inductors, ceramic capacitors, or varicap diodes, even
ordinary power diodes. You can get the risetime to go down as a step
propagates down the line.


Check; i have heard stories about sub-nanosecond risetimes.

Tim Williams wrote:
"John Larkin" wrote in
message ...
Nonlinear transmission lines are cool. You can make them with
saturating inductors, ceramic capacitors, or varicap diodes, even
ordinary power diodes. You can get the risetime to go down as a step
propagates down the line.

I've been tempted to try that before -- a necklace of ferrite beads
might run you only a couple bucks per foot, and with the velocity factor
so low, that's enough to get something snappy with jellybean transistors
(a few nanoseconds' input edge). Downside is the impedance, which must
be in the kohms range (small signal). Hard to get much current into it
like that. Supposedly, ferrite has a high e_r as well, so the impedance
probably isn't *that* bad, but it's hard to ensure a consistent airgap
(or lack thereof) with loose beads.

I expect the pulse width is limited as much by geometry (ferrite beads
being a coax structure, periodically disturbed by the imperfect faces
between beads and imperfect spacing to the conductors) as by material
dispersion (losses, etc).
* The idea is (i think the proper term is) lumped constant; transmission
line = = L-C-L-C-L etc, in this case series L and shunt C.
So use small (size) L (small beads) with small C (chip caps) to
ground plane. The caps also being the mechanical "standoff" for the wire
used.


The not-even-ludicrous speed[1] generators do it with schottky junctions
in InP (like the LeCroy 200GHz scope) IIRC. With harmonics near the THz,
there isn't any physical length to spare before the signal dissipates,
anyway.
* Dissipate? I think not. HF part attenuated tho is something to look at.


Tim

[1]
"Aah! What the hell was that?!"
"Spaceball 1!"
"They've gone to plaid!"

"Robert Baer" wrote in message
net...
Try amorphous or nanocrystalline cores, too. They switch quite quickly.
I've built a toy circuit that generated enough harmonics to cause itself
to ring a bit; the risetime was under 1us.

Where can i buy some?
Are there various sizes available?

You can find a small one in any computer PSU: they use a saturable core
(square B-H curve) to regulate the 3.3V supply off the 5V winding. (Thanks
to separate regulation, and the high gain of the TL431, the 3.3V supply has
substantially better regulation than the 5 or 12V rails!)

They're sometimes used in line filters too, where they offer unusually low
cutoff frequencies. The average COTS line filter peaks around 10MHz;
amorphous/nano cores peak around 0.1-1MHz instead. Look for unusually heavy
epoxy or plastic-cased cores; occasionally the plastic case cores hold a
Finemet or ferrite toroid, so peek inside if you can open it easily.

Most core suppliers have them; Elna Magnetics carries VAC, Adams Magnetic
carries Metglas/HMG. Check also with any local power control reps, might
get some samples or something. There's also a few sitting he
http://www.surplussales.com/Inductor...FerToro-3.html
I picked up a few of the 50B12-1Ds (square permalloy IIRC), which I would
guess might switch 100W each if you push 'em. I ran a few watts through one
as a test:
http://myweb.msoe.edu/williamstm/Ima...g_Amp_Test.png

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

On Wed, 13 Jun 2012 23:25:25 -0800, Robert Baer
wrote:

John Larkin wrote:
On Sun, 10 Jun 2012 11:25:54 -0500, "Tim Williams"
wrote:

Looks quite good, actually. Like Fred says, integrate, then you will have
it right.

Try amorphous or nanocrystalline cores, too. They switch quite quickly.
I've built a toy circuit that generated enough harmonics to cause itself to
ring a bit; the risetime was under 1us.

Physics guys use the stuff to generate fast (1us) current pulses for
particle accelerators and junk. They drive them with 3kA+ IGBTs, which
don't even switch as fast (~1us). Lots of power density.

Tim

Nonlinear transmission lines are cool. You can make them with
saturating inductors, ceramic capacitors, or varicap diodes, even
ordinary power diodes. You can get the risetime to go down as a step
propagates down the line.


Check; i have heard stories about sub-nanosecond risetimes.

Robert, please measure again with primary current on the x axis and the
integral of secondary voltage on the y axis. I know what i expect to see
and why. BTW you may just measure secondary voltage and numerically
integrate it later, just so that it is easier to make the measurements.

?-)

On Thu, 14 Jun 2012 18:45:50 -0500, "Tim Williams"
wrote:

"Robert Baer" wrote in message
lnet...
Try amorphous or nanocrystalline cores, too. They switch quite quickly.
I've built a toy circuit that generated enough harmonics to cause itself
to ring a bit; the risetime was under 1us.

Where can i buy some?
Are there various sizes available?

You can find a small one in any computer PSU: they use a saturable core
(square B-H curve) to regulate the 3.3V supply off the 5V winding. (Thanks
to separate regulation, and the high gain of the TL431, the 3.3V supply has
substantially better regulation than the 5 or 12V rails!)

I have an issue he Square loop (square B-H curve) is not the same as
saturable core at all.

Square B-H curve:
http://info.ee.surrey.ac.uk/Workshop...t/results.html

Saturable reactor (and magnetic amplifier):
http://www.tpub.com/neets/book8/32m.htm
http://www.toshiba.com/taec/components/Generic/M1.pdf
The best saturable reactors / magnetic amplifiers want trapezoidal loops
with significant slant on the right and left sides.

@Robert
Simple circuit to plot B-H curves:
http://info.ee.surrey.ac.uk/Workshop...Ckt/index.html

They're sometimes used in line filters too, where they offer unusually low
cutoff frequencies. The average COTS line filter peaks around 10MHz;
amorphous/nano cores peak around 0.1-1MHz instead. Look for unusually heavy
epoxy or plastic-cased cores; occasionally the plastic case cores hold a
Finemet or ferrite toroid, so peek inside if you can open it easily.

Most core suppliers have them; Elna Magnetics carries VAC, Adams Magnetic
carries Metglas/HMG. Check also with any local power control reps, might
get some samples or something. There's also a few sitting he
http://www.surplussales.com/Inductor...FerToro-3.html
I picked up a few of the 50B12-1Ds (square permalloy IIRC), which I would
guess might switch 100W each if you push 'em. I ran a few watts through one
as a test:
http://myweb.msoe.edu/williamstm/Ima...g_Amp_Test.png

Tim

On Fri, 29 Jun 2012 19:44:08 -0700, josephkk
wrote:

On Thu, 14 Jun 2012 18:45:50 -0500, "Tim Williams"
wrote:

"Robert Baer" wrote in message
alnet...
Try amorphous or nanocrystalline cores, too. They switch quite quickly.
I've built a toy circuit that generated enough harmonics to cause itself
to ring a bit; the risetime was under 1us.

Where can i buy some?
Are there various sizes available?

You can find a small one in any computer PSU: they use a saturable core
(square B-H curve) to regulate the 3.3V supply off the 5V winding. (Thanks
to separate regulation, and the high gain of the TL431, the 3.3V supply has
substantially better regulation than the 5 or 12V rails!)

I have an issue he Square loop (square B-H curve) is not the same as
saturable core at all.

Square B-H curve:
http://info.ee.surrey.ac.uk/Workshop...t/results.html

Saturable reactor (and magnetic amplifier):
http://www.tpub.com/neets/book8/32m.htm
http://www.toshiba.com/taec/components/Generic/M1.pdf
The best saturable reactors / magnetic amplifiers want trapezoidal loops
with significant slant on the right and left sides.

OOps make that rhombic (~parallel sides).

@Robert
Simple circuit to plot B-H curves:
http://info.ee.surrey.ac.uk/Workshop...Ckt/index.html

They're sometimes used in line filters too, where they offer unusually low
cutoff frequencies. The average COTS line filter peaks around 10MHz;
amorphous/nano cores peak around 0.1-1MHz instead. Look for unusually heavy
epoxy or plastic-cased cores; occasionally the plastic case cores hold a
Finemet or ferrite toroid, so peek inside if you can open it easily.

Most core suppliers have them; Elna Magnetics carries VAC, Adams Magnetic
carries Metglas/HMG. Check also with any local power control reps, might
get some samples or something. There's also a few sitting he
http://www.surplussales.com/Inductor...FerToro-3.html
I picked up a few of the 50B12-1Ds (square permalloy IIRC), which I would
guess might switch 100W each if you push 'em. I ran a few watts through one
as a test:
http://myweb.msoe.edu/williamstm/Ima...g_Amp_Test.png

Tim

"josephkk" wrote in message
...
You can find a small one in any computer PSU: they use a saturable core
(square B-H curve) to regulate the 3.3V supply off the 5V winding.
(Thanks
to separate regulation, and the high gain of the TL431, the 3.3V supply
has
substantially better regulation than the 5 or 12V rails!)

I have an issue he Square loop (square B-H curve) is not the same as
saturable core at all.


Well, square and narrow. Since NdFeB makes a bad reactor..

The classical "saturable reactor" consists of two magnetic paths and three
windings, imposing a CM-diff magnetic field between the two paths (usually
implemented as two toroids or the two arms of an E-core). This isn't to say
you can't get the same behavior (magnetic amplification, saturable reactor,
etc.) from a single winding, it's just arranged differently, and in
particular, diodes are used to steer the waveform as required.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

"Fred Abse" wrote in message
news
Secondly, permeability may not be independent of frequency, in some
ferrites. A glance at published data will illustrate this.

Even just playing with an iron-cored transformer can give one an interesting
perspective. Apply square wave, variable frequency; measure current. At
low frequencies, the current waveform is a sloping triangle, due to
inductive integration. Fine and dandy. Increase the frequency: the slope
gets less and less (as a percentage of amplitude), and the resistive (eddy
current) component dominates. Now increase the frequency still further.
Instead of falling or remaining constant, the current amplitude will
actually increase, because less and less core is being used (the skin depth
is significantly thinner than the laminations) and the effective
permeability (which at this point is almost entirely imaginary, mu'') is
falling.

Hard to see the same in ferrite, but the same effects apply. Different
mechanisms, same effects.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

Tim Williams wrote:
"Fred Abse" wrote in message
news
Secondly, permeability may not be independent of frequency, in some
ferrites. A glance at published data will illustrate this.

Even just playing with an iron-cored transformer can give one an
interesting perspective. Apply square wave, variable frequency; measure
current. At low frequencies, the current waveform is a sloping triangle,
due to inductive integration. Fine and dandy. Increase the frequency:
the slope gets less and less (as a percentage of amplitude), and the
resistive (eddy current) component dominates. Now increase the frequency
still further. Instead of falling or remaining constant, the current
amplitude will actually increase, because less and less core is being
used (the skin depth is significantly thinner than the laminations) and
the effective permeability (which at this point is almost entirely
imaginary, mu'') is falling.

Hard to see the same in ferrite, but the same effects apply. Different
mechanisms, same effects.

Tim

All of the cores i played with so far are/shouldbe square loop or
near that, as they are all from computers like 7090 era; if i remember
right, one or two are from the IBM 1620.
The 1620 had variations, one a "scratchpad" array for multiplication
tables; another was an index register scratchpad.
Interesting idea i had at the time (probably in use then) was to run
X-Y drive into a core array, each core having an output line then used
as a select line to the actual memory array; would require two "drive"
arrays, one for X and one for Y.
Shoot, even the G-15 magnetic drum-based computer had a small core
array - i think mainly for an index register.