Ferrite desaturation in slow motion

On Sunday, September 27, 2020 at 11:43:17 AM UTC-7, Piotr Wyderski wrote:

> It led me to a follow-up question: how do ferrite losses depend on 
> saturation? If the frequency is sufficiently high, say ~1GHz, and the 
> wire passes through the core, can I turn on/off the losses by saturating 
> the ferrite? Say the attenuation range of interest is 2:1 or more. 

The usual picture of grain magnetization has small domains which grow and
shrink.   A true saturation would make the whole grain ONE domain, and
coming out of saturation means generating new domain boundaries, not just
moving preexisting ones; that's a spontaneous symmetry break, and it always
means entropy, i.e. energy loss to heat.  Saturation might depend on grain sizes.

Tim Williams wrote:

> "Lossless" is a physical impossibility, the real question is how much 
> loss can you tolerate, under what conditions?

"Lossless" = not introducing much additional losses beyond what already 
is being dissipated. That limits the methods available to magnetic field 
sensing and busbar voltage drop sensing. 100 amps now, 500A in the 
future, low voltage, loss budget is 200mW. I want to detect that there 
is no significant current flowing, I don't care how high the current is
if it is beyond some threshold. Quite a similar problem to synchronous 
rectification.

The sensor needs to be relatively small, fast and reliable and cost 
under 30 dollars in low volume. Hence my musing about saturable 
ferrites, which have all the required properties, perhaps except for the 
speed: 10ns is my pain threshold, 5ns would be perfect.

> Notice it's not cheating to use

No problems with that.

	Best regards, Piotr

Bus bar?  So, this thing is big?

What characteristic impedance is the bus bar?  Low SWR?  If it's not known 
with confidence this is a meaningless measurement.

This is where directional couplers take over, with good reason.

Tim

-- 
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Design
Website: https://www.seventransistorlabs.com/

"Piotr Wyderski" <peter.pan@neverland.mil> wrote in message 
news:rl3ri5$2t55u$1@portraits.wsisiz.edu.pl...
> Tim Williams wrote:
>
>> "Lossless" is a physical impossibility, the real question is how much 
>> loss can you tolerate, under what conditions?
>
> "Lossless" = not introducing much additional losses beyond what already is 
> being dissipated. That limits the methods available to magnetic field 
> sensing and busbar voltage drop sensing. 100 amps now, 500A in the future, 
> low voltage, loss budget is 200mW. I want to detect that there is no 
> significant current flowing, I don't care how high the current is
> if it is beyond some threshold. Quite a similar problem to synchronous 
> rectification.
>
> The sensor needs to be relatively small, fast and reliable and cost under 
> 30 dollars in low volume. Hence my musing about saturable ferrites, which 
> have all the required properties, perhaps except for the speed: 10ns is my 
> pain threshold, 5ns would be perfect.
>
>> Notice it's not cheating to use
>
> No problems with that.
>
> Best regards, Piotr 

Tim Williams wrote:

> Bus bar?� So, this thing is big?

The busbar is 14cm long, 3x16mm copper slab. Dunno if that's big.

> What characteristic impedance is the bus bar?� Low SWR?� If it's not 
> known with confidence this is a meaningless measurement.

I can measure that, but I am not sure the wave properties are relevant 
at this scale.

> This is where directional couplers take over, with good reason.

I was referring to microwaves just because a saturable reactor sampled 
at 1GHz would provide me with 1ns resolution (or 500ps, in fact, if I 
use both halves of the sampling waveform). But the ferrite would need to 
be comparably fast. For example, switching a 14mm 3R1 toroid takes about 
a microsecond, which would be 1e3 times too slow. I have various core 
memory rings from USSR and DDR factories, but I don't expect them to go 
beyond 10MHz. Don't ask me how to pass 500 amps through a 1mm ID core, 
though... My googling shows that if ferrite needs to be fast, then that 
would solely mean the gyromagnetic effects known from the lithium-doped 
microwave ferrites. That is what Bill Sloman was referring to. Doable in 
principle, but the complexity would go through the roof at supersonic 
speed � apparently, such a beautiful idea killed by such an ugly fact.

	Best regards, Piotr

On Friday, October 2, 2020 at 1:30:29 AM UTC+10, Piotr Wyderski wrote:
> Tim Williams wrote: 
> 
> > Bus bar? So, this thing is big?
> The busbar is 14cm long, 3x16mm copper slab. Dunno if that's big.
> > What characteristic impedance is the bus bar? Low SWR? If it's not 
> > known with confidence this is a meaningless measurement.
> I can measure that, but I am not sure the wave properties are relevant 
> at this scale.
> > This is where directional couplers take over, with good reason.
> I was referring to microwaves just because a saturable reactor sampled 
> at 1GHz would provide me with 1ns resolution (or 500ps, in fact, if I 
> use both halves of the sampling waveform). But the ferrite would need to 
> be comparably fast. For example, switching a 14mm 3R1 toroid takes about 
> a microsecond, which would be 1e3 times too slow. I have various core 
> memory rings from USSR and DDR factories, but I don't expect them to go 
> beyond 10MHz. Don't ask me how to pass 500 amps through a 1mm ID core, 
> though... My googling shows that if ferrite needs to be fast, then that 
> would solely mean the gyromagnetic effects known from the lithium-doped 
> microwave ferrites. That is what Bill Sloman was referring to. 

Sadly, no. I was just talking about isolated transition metal atoms in chemical compounds.

If you wanted to measure a magnetic field fast, a spectroscopic approach might do it. You can get magnetic-field-dependent splitting of  electronic absorbtion spectra.

https://www.scielo.br/scielo.php?script=sci_arttext&pid=S0103-97332004000400044

Turning that into a scheme for measuring current in 14cm long 1.6cm wide , 3mm thick copper bus bar could be tricky. The Brazilians used a tungsten lamp as their light source, but the molecular beam deposition scheme to get the light-absorbing sample would be more difficult to duplicate.

> Doable in principle, but the complexity would go through the roof at supersonic  speed — apparently, such a beautiful idea killed by such an ugly fact. 

It is a beautiful idea . I wish that I'd had it.

-- 
Bill Sloman, Sydney

"Piotr Wyderski" <peter.pan@neverland.mil> wrote in message 
news:rl4sme$37732$1@portraits.wsisiz.edu.pl...
> The busbar is 14cm long, 3x16mm copper slab. Dunno if that's big.
>

Ok... what else is around it?  Currents don't just show up out of nowhere, 
definitely not at these time scales.

A figure for Zo is basically answering this quantitatively.

What's making the current (or the lack of it), anyway?  Why is the time 
scale so short?  Why is current the best way to measure it, why not voltage 
or whatever?

(You gave synchronous rectification as an example application, but the 
scales aren't commensurate, based on what I've heard.  For example, a ~100A 
bus bar at mains frequency, could take dozens of microseconds to commutate, 
who cares.  A 100A bus bar strung between IGBT modules would be very much 
that size, but still not the rate, 100ns being adequate.  And neither would 
seem to require the precision asked for?)


>> What characteristic impedance is the bus bar? Low SWR? If it's not known 
>> with confidence this is a meaningless measurement.
>
> I can measure that, but I am not sure the wave properties are relevant at 
> this scale.

Not quite, no, but you're in the cutoff region where, depending on system Z 
vs. Zo, either wave properties or their LF equivalents are mandatory to 
consider.  Put another way, you cannot design such a system using statics.

Put still another way -- current transformers for instance don't depend on 
ferrites at all, they're just improved with them.  Apparently being in the 
HF cutoff region, the core hardly matters at all, and it's all about winding 
geometry.  You can still play with magnetics, sure, but you aren't going to 
be able to

Or like, what's the repeat rate?  You said 200mW (which is a tiny, TINY 
fraction of anything you'll notice from that bus bar; who's asking, the 
CIA?..), but the de/magnetization of even a small amount of ferrite, at more 
than a modest repeat rate, will consume that easily.  (The current 
transformer, given most likely geometries, will probably be an order of 
magnitude higher, or even more.)

Tim

-- 
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Design
Website: https://www.seventransistorlabs.com/ 

Flyguy wrote:
> On Sunday, September 27, 2020 at 11:28:00 PM UTC-7, Piotr Wyderski wrote:
>> George Herold wrote:
>>
>>> One purpose of the over wrapped coil was to cancel the
>>> Earth's B-field... so that gives you some estimate of the
>>> fields involved.
>>
>> Yes, fluxgates can be sensitive and accurate, but they are pretty slow.
>> The BW of those I know of is <1MHz. Here I have quite the opposite
>> problem: accuracy can be low, and no linearity is required (a magnetic
>> window comparator is what I need), but the time scale of H change is on
>> the order of 10ns. I am trying to figure out what the B change would
>> then be.
>>
>> 	Best regards, Piotr
> 
> Here is a magnetic field sensor with a bandwidth of 200 MHz:
> https://tinyurl.com/y364ubhy
> 
<Error><Code>AccessDenied</Code><Message>Access denied</Message></Error>

Tim Williams wrote:
> "Piotr Wyderski" <peter.pan@neverland.mil> wrote in message 
> news:rl4sme$37732$1@portraits.wsisiz.edu.pl...
>> The busbar is 14cm long, 3x16mm copper slab. Dunno if that's big.
>>
> 
> Ok... what else is around it?  Currents don't just show up out of 
> nowhere, definitely not at these time scales.
> 
[...]

Current transformers for ~100A at ns timescales? That's the
sort of device I make for particle accelerators. Big expensive
things with coaxial geometries. There are two types of ferrite
in them: Big rings with highish permeability to extend the
frequency response down to about 100kHz, and absorbing tiles
so that cavity resonances don't spoil the high frequencies
too much. At high frequency, the rings might as well not be
there.

I get sub-100ps risetimes, or equivalently, bandwidths of the
order of 4GHz. Here's a picture of two of them in CERN's PS:
<http://psring.web.cern.ch/psring/showpicture.php?section=3>.

Jeroen Belleman

Tim Williams wrote:

> Ok... what else is around it?� Currents don't just show up out of 
> nowhere, definitely not at these time scales.

It depends on the pulse shape that is interesting at the moment.
It can be a transformer as well as a capacitor bank. Here is an array of 
~700 10uF MLCCs in parallel, arranged vertically and soldered between 
two parallel copper plates. The low inductance porn before soldering:

https://i.postimg.cc/y651xS71/cap-array.png

> What's making the current (or the lack of it), anyway?� Why is the time 
> scale so short?� Why is current the best way to measure it, why not 
> voltage or whatever?

It is a setup for measuring current pulse response of various components 
and modules. SiC half-bridge modules, synchronous rectifiers and so on.
Currently, I am interested in "ideal" SR driving for HF resonant 
converters. Apparently, nobody knows how to drive SR in an LLC converter 
close to the theoretic optimum. There are various approximate 
approaches, but at the moment, I consider them to be all failures. And 
if semiconductor electronics fails to provide a solution, I wanted to 
join the game with magnetic switches and see what happens. The critical 
part is I_OFF sensing with low phase error.

> (You gave synchronous rectification as an example application, but the 
> scales aren't commensurate, based on what I've heard.� For example, a 
> ~100A bus bar at mains frequency, could take dozens of microseconds to 
> commutate, who cares. 

For the mains frequency, I already have a very robust and super cheap 
saturable reactor-based solution, but you know that. Say I want to speed 
it up by a tiny factor of 1e6.

> You can still play with magnetics, sure, but you aren't going to be able to

Not in the usual way at least, as it appears. But there are still those 
spin resonance effects that seem to be precisely at the right time 
scale. A YIG oscillator is based on a very arcane physics and 
manufacturing technology, but at the end of the day the entire 
construction boils down to two wires with a grain of sand sitting in 
between. Now *that* is simplicity. So why not at least consider using 
this simplicity for SR drive? It does not even need to be an oscillator, 
a current-controlled attenuator or phase shifter would perfectly 
suffice. Simple, small, very robust and blazing fast.

Moreover, you can easily adjust the switching threshold with an IDAC 
powering a coil -- if di/dt is more or less known, you could hit the 
true zero crossing quite accurately, compensating all the delays. But do 
I need special ferrites from the Ministry of Magic or would a piece of 
NiZn suffice? I have no idea. So this would be the picture at the pipe 
dream level, now I am about to enter the crash with the reality phase. 
Probably sorrow comes next, but I'll never know without trying.

As a matter of fact, this compensating current idea was the target 
approach from the very beginning, but the working principle was simpler:

1. use a square hysteresis loop tiny ferrite core,
2. pre-magnetise it in the direction that is opposing the main current flow,
3. turn on the main current and record the core switching events.

Sort of the coincident current core memory principle running backwards. 
Sadly, sir, the core made by the commies is too slow.

> Or like, what's the repeat rate?

This is for HF converters, so in the 1-10MHz range.

> You said 200mW (which is a tiny, TINY fraction of anything you'll notice from that bus bar

There is just no point in using a lossy SR, especially at low loads. Use 
a Schottky and get over the heating. So I have decided to put the 
acceptance threshold at about the gate drive power level. 200mW is a lot 
of power.

> but the de/magnetization of even a small amount of ferrite, at 
> more than a modest repeat rate, will consume that easily.

That can very well be the case.

	Best regards, Piotr

Jeroen Belleman wrote:

> Current transformers for ~100A at ns timescales? That's the
> sort of device I make for particle accelerators. Big expensive
> things with coaxial geometries. 

Your "coaxial CT" resulted in a nice finding:

https://www.researchgate.net/publication/3170961_Coaxial_current_transformer_for_test_and_characterization_of_high-power_semiconductor_devices_under_hard_and_soft_switching

And their application is very similar to mine. Thanks a lot!

	Best regards, Piotr